|
On July 07 2025 23:51 johnnyh123 wrote:
Honestly, the competitiveness in SC2 between today and the early 2010s is, in my opinion, the difference between State HS football vs NCAA D1. I'd put the number somewhere around 10x difference.
The typical peak age of esports players is the early-20s. My question is, who, amongst anyone that is reputable today, is 21 (US drinking age) or younger? I don't know any today. And i've watched every single premier tournament since Day 1 in SC2 and played. In early-2010s, literally half the pros are 21 or younger (unfortunatelly, including myself, and not anymore).
I think it makes sense to have an era-multiplier, the real question is how do we assign the #, rather than whether we should. Afterall, any serious SC2 professional, analyst, or fan, would 100% know, today's SC2 is nowhere near as competitive as years ago. What this number is, I can't objectively tell you, but subjectively, probably somewhere around 10x compared to the most competitive era back then.
Back then, things were exciting, new up and coming players are popping up left and right. Now, I can easily look at the list, and tell you exactly the name of the few players that will win, and their likely % of winning.
In which esport is the best player right now 21 and under?
It's not SC2, not BW, not CS2 (though you could at least make a case for it). Don't know about the other esports.
|
Northern Ireland25283 Posts
On July 08 2025 16:40 Harris1st wrote:Show nested quote +On July 07 2025 23:51 johnnyh123 wrote:
Honestly, the competitiveness in SC2 between today and the early 2010s is, in my opinion, the difference between State HS football vs NCAA D1. I'd put the number somewhere around 10x difference.
The typical peak age of esports players is the early-20s. My question is, who, amongst anyone that is reputable today, is 21 (US drinking age) or younger? I don't know any today. And i've watched every single premier tournament since Day 1 in SC2 and played. In early-2010s, literally half the pros are 21 or younger (unfortunatelly, including myself, and not anymore).
I think it makes sense to have an era-multiplier, the real question is how do we assign the #, rather than whether we should. Afterall, any serious SC2 professional, analyst, or fan, would 100% know, today's SC2 is nowhere near as competitive as years ago. What this number is, I can't objectively tell you, but subjectively, probably somewhere around 10x compared to the most competitive era back then.
Back then, things were exciting, new up and coming players are popping up left and right. Now, I can easily look at the list, and tell you exactly the name of the few players that will win, and their likely % of winning. In which esport is the best player right now 21 and under? It's not SC2, not BW, not CS2 (though you could at least make a case for it). Don't know about the other esports.
For me it’s a myth of sorts, as eSports are new and evolving. Great players tend to be young, but I don’t think they remotely have to be.
Player A is pushing the boundaries of what’s possible, and dominates for a bit. Next generation push it further as the game is more understood, and player B is the new dominant player. Player A may have been player B if they came later, but it’s hard to effectively push to that new level and relearn a lot.
Whereas if a game remains stable for years, it seems older players do remain very competitive
Also unless you’re really a top player, it’s not a great earner. What’s an OK lifestyle in your teens or early 20s isn’t necessarily when you’re older, so there’s attrition there as well
|
On July 08 2025 11:36 rwala wrote:Show nested quote +On July 05 2025 14:56 PremoBeats wrote:On July 05 2025 00:00 Balnazza wrote:But while herO has now won more money than any other protoss I do not believe he is the protoss GOAT. Which gives you two options: Either prizemoney is a bad option to calculate the GOAT or you specifically start manipulating the statistic until it gives you the "correct" answer...but then why have a statistic in the first place? The entire era- and balance-weighting is a flawed concept to begin with considering how subjective it is, it really does not get improved by also "punishing" big prizepools, which usually correspond with important titles. Also...why did you disqualify WCS Europe/America? WCS Europe 2014 Season 1, in which MMA placed 2nd, fulfills all your criteria: Offline (mostly), not an invitational (same format as GSL), Korean players (you said korean players OR korean qualifier, not both) and the prizepool is higher than 2K. Are you really taking this list seriously? There is a 4 times multiplier on era, which no sound mathemetical approach would suggest. Players get penalized for solo carrying their race. The obvious language used... like what :D And I can't decide... are these last two posts written by a person trying to troll hype or ridicule this thread, or are they actually bot accounts programmed to do that? The first text is definitely written by AI. It’s honestly no different than yours at a conceptual level
You are not entirely wrong at the highest level of abstraction. But from any meaningful, analytical, methodological or communicative level, these two articles are clearly not the same.
One is a tongue-in-cheek rhetorical piece that makes you think about balance and criteria, while delivering truths about GOAT-analyses, while the other is a rigorous analysis trying to quantify properly. While one is trollish, sarcastic and borderline mocking, the other stays more neutral and carefully reasoned. Saying these two are conceptually the same ignores the core difference in methodology, purpose and seriousness. It is like you are saying that a scientific study and a satirical cartoon about climate change are conceptually the same, as they both address climate change.
On July 08 2025 16:03 MJG wrote:Show nested quote +On July 05 2025 14:56 PremoBeats wrote:On July 05 2025 00:00 Balnazza wrote:But while herO has now won more money than any other protoss I do not believe he is the protoss GOAT. Which gives you two options: Either prizemoney is a bad option to calculate the GOAT or you specifically start manipulating the statistic until it gives you the "correct" answer...but then why have a statistic in the first place? The entire era- and balance-weighting is a flawed concept to begin with considering how subjective it is, it really does not get improved by also "punishing" big prizepools, which usually correspond with important titles. Also...why did you disqualify WCS Europe/America? WCS Europe 2014 Season 1, in which MMA placed 2nd, fulfills all your criteria: Offline (mostly), not an invitational (same format as GSL), Korean players (you said korean players OR korean qualifier, not both) and the prizepool is higher than 2K. Are you really taking this list seriously? There is a 4 times multiplier on era, which no sound mathemetical approach would suggest. Players get penalized for solo carrying their race. The obvious language used... like what :D And I can't decide... are these last two posts written by a person trying to troll hype or ridicule this thread, or are they actually bot accounts programmed to do that? The first text is definitely written by AI. I take it just as seriously as any other list that thinks WoL, HotS and LotV are comparable games that can have a unified GOAT, or that Protoss, Terran and Zerg are comparable races that can have a unified GOAT, so I guess I take it just as seriously as your list.  At least Miz went to the effort of writing some genuinely entertaining articles to go alongside their rating! And yes, those posts are clearly by bots and should be reported. Have you not noticed that bots are becoming a big problem around here? Damn varmints.
Fair enough... I warned everyone beforehand, that my approach is dry and not very entertainingly written 
|
On July 05 2025 00:00 Balnazza wrote:Show nested quote +But while herO has now won more money than any other protoss I do not believe he is the protoss GOAT. Which gives you two options: Either prizemoney is a bad option to calculate the GOAT or you specifically start manipulating the statistic until it gives you the "correct" answer...but then why have a statistic in the first place? The entire era- and balance-weighting is a flawed concept to begin with considering how subjective it is, it really does not get improved by also "punishing" big prizepools, which usually correspond with important titles. Also...why did you disqualify WCS Europe/America? WCS Europe 2014 Season 1, in which MMA placed 2nd, fulfills all your criteria: Offline (mostly), not an invitational (same format as GSL), Korean players (you said korean players OR korean qualifier, not both) and the prizepool is higher than 2K.
I grant you that prize winnings isn't the best criteria to use to find your GOAT, but then, I never claimed this to be the case. What my list offers and the reason I started doing this more lazy list, is that it does not put Rain as #10 for some reason, and dark outside of top 10 initially, it does not tell you that serral is 4x better than Rogue. While it's hard on serral for simply not having spawned in korea and players that fled from korea during the toughest era, and is tough on invitationals such as gsl vs. the world or online katowice, it actually paints a pretty, pretty picture of what the GOAT ranking should look like, non-koreans might be slightly misplaced, and the resolution is not as fine as some of these other approaches, but at the same time, it doesn't wank itself into a corner by bias after bias. Here, I let tournaments speak for themselves, herO gets nada from his 2nd place at katowice because that's how the tournament saw it, I'm not layering abstraction upon abstraction and somehow concluding that current GSL still finds itself on the highest pedestal, despite the obvious downsizing. The reason I ventured on this endeavour was that using top 10 on prize earnings gives a better insight into who is the greatest player than all of these artificial lists, not because it's insightful, but because the lists are bad. So, I thought why not use prize winnings and fine tune it, and along the way I've found some neat discoveries. Before this whole debate I was actually on the serral camp and you can go back and read my older posts, I wanted to look at this to educate myself so that I could write more thoughtful posts in the debate. I'm satisfied with the current list and find it way better than the rest, hence the name goat ranking of goat rankings, and this is despite using prize winnings, the bad criteria. Actually, it's not so far from stuchiu's old ranking, if I cut off from 2015, and that one was received rly well, other than half the readers wanting Life at 1st, and feeling parting is too low. But considering that I take into account balance and that great wins were still ahead that year for many protosses and Life, I do find my list congruent with that ranking. The mizen list wouldn't agree with that one, because only mvp and rain appear suddenly and out of the blue, but nowhere're mc, taeja and the other terrans to be seen.. I respect the research, and I'm sure both of you have learned something from the research, but as far as the lists goes, they're bad.
Most ppl would agree that if one player wins every tournament within a year, but only gets 2nd at the biggest event that that player is greater than the winner of said biggest event despite being down on a few dollars.
WCS eu/am wasn't open for ALL koreans via qualifiers.
|
but at the same time, it doesn't wank itself into a corner by bias after bias
This is literally what you did though? Just to take the WCS eu/am again, it completly fulfills your criteria and yet you disregard it. Yes, the qualifier was open to ALL koreans, they just had to move to Europe or America. That is essentially the same requirement GSL had, you had to live in Korea for a while. The biggest reason not more koreans did it was most likely Proleague, because you couldn't really play Proleague when you played WCS. But that isn't an excluding factor.
This was pre-regionlock, when koreans were actually forced to be residents of Europe or America. But the whole criticism before that was how lackluster some koreans were in "their" new region, how little they contributed to the regional scenes.
If you want to take prizemoney, just take prizemoney. And yes, then you will have to live with the fact that Clem (by winning EWC), Reynor (by winning Gamers8) and even Maru to a degree (by winning WESG) have a bit of inflated stats, but that is the hand you have been dealt with if you pick prizemoney as the defining statistic. But the second you introduced your "fine-tuning", you made yourself vulnerable...and as you notice, a lot of people (if not all of them) don't agree with your fine-tuning, not even the ones that agree with the outcome. Because it is such a biased and random bunch of multipliers.
Just as an example I will point out again the HSC I mentioned earlier, that had barely anyone of notice in attendance but gave out more points than Katowice the next year, just because both tournaments were in different "eras"...divided by two months. damn did the scene drop in two months! And don't get me started on the highly subjective "imbalance"-multiplier...
No GOAT list will ever be perfect, that's impossible. If we had like 40 years of SC2 that would have been somewhat the same we would have a good metric to decide by, but considering how much shift and change there has been in the ~15 years lifespan of this game, it will always come down to what statistic you highlight and what metric you choose.
But sadly, your way just wasn't solid at all. It was highly biased and too radical to correctly adjust for change (not that you particularly need to).
But the funniest part is honestly that you think your approach is completly unbiased...
|
On July 12 2025 06:50 Balnazza wrote:Show nested quote + but at the same time, it doesn't wank itself into a corner by bias after bias This is literally what you did though? Just to take the WCS eu/am again, it completly fulfills your criteria and yet you disregard it. Yes, the qualifier was open to ALL koreans, they just had to move to Europe or America. That is essentially the same requirement GSL had, you had to live in Korea for a while. The biggest reason not more koreans did it was most likely Proleague, because you couldn't really play Proleague when you played WCS. But that isn't an excluding factor. This was pre-regionlock, when koreans were actually forced to be residents of Europe or America. But the whole criticism before that was how lackluster some koreans were in "their" new region, how little they contributed to the regional scenes. If you want to take prizemoney, just take prizemoney. And yes, then you will have to live with the fact that Clem (by winning EWC), Reynor (by winning Gamers8) and even Maru to a degree (by winning WESG) have a bit of inflated stats, but that is the hand you have been dealt with if you pick prizemoney as the defining statistic. But the second you introduced your "fine-tuning", you made yourself vulnerable...and as you notice, a lot of people (if not all of them) don't agree with your fine-tuning, not even the ones that agree with the outcome. Because it is such a biased and random bunch of multipliers. Just as an example I will point out again the HSC I mentioned earlier, that had barely anyone of notice in attendance but gave out more points than Katowice the next year, just because both tournaments were in different "eras"...divided by two months. damn did the scene drop in two months! And don't get me started on the highly subjective "imbalance"-multiplier... No GOAT list will ever be perfect, that's impossible. If we had like 40 years of SC2 that would have been somewhat the same we would have a good metric to decide by, but considering how much shift and change there has been in the ~15 years lifespan of this game, it will always come down to what statistic you highlight and what metric you choose. But sadly, your way just wasn't solid at all. It was highly biased and too radical to correctly adjust for change (not that you particularly need to). But the funniest part is honestly that you think your approach is completly unbiased...
Yeah, this measurement is quite faulty I think, but it brings a great point across which ironically is the one OP quoted in his post
"If you torture the data long enough, it will confess to anything"
At the end of the day you can use all the fancy stats and data you want, but it doesn't make your ranking any less subjective than someone's guts feeling, whether it's premobeats way overvaluing things like winrates or aligulac rating to prob Serral or the thing ejozl did here.
And we all know the one true Goat ranking is this one:
1. Serral
2. Maru (extremely close to Serral)
3. Inno
4. Rogue
5. Life
6. Zest
7. sOs
8. herO
9. Dark
10. Stats
|
On July 13 2025 05:44 Charoisaur wrote:Show nested quote +On July 12 2025 06:50 Balnazza wrote: but at the same time, it doesn't wank itself into a corner by bias after bias This is literally what you did though? Just to take the WCS eu/am again, it completly fulfills your criteria and yet you disregard it. Yes, the qualifier was open to ALL koreans, they just had to move to Europe or America. That is essentially the same requirement GSL had, you had to live in Korea for a while. The biggest reason not more koreans did it was most likely Proleague, because you couldn't really play Proleague when you played WCS. But that isn't an excluding factor. This was pre-regionlock, when koreans were actually forced to be residents of Europe or America. But the whole criticism before that was how lackluster some koreans were in "their" new region, how little they contributed to the regional scenes. If you want to take prizemoney, just take prizemoney. And yes, then you will have to live with the fact that Clem (by winning EWC), Reynor (by winning Gamers8) and even Maru to a degree (by winning WESG) have a bit of inflated stats, but that is the hand you have been dealt with if you pick prizemoney as the defining statistic. But the second you introduced your "fine-tuning", you made yourself vulnerable...and as you notice, a lot of people (if not all of them) don't agree with your fine-tuning, not even the ones that agree with the outcome. Because it is such a biased and random bunch of multipliers. Just as an example I will point out again the HSC I mentioned earlier, that had barely anyone of notice in attendance but gave out more points than Katowice the next year, just because both tournaments were in different "eras"...divided by two months. damn did the scene drop in two months! And don't get me started on the highly subjective "imbalance"-multiplier... No GOAT list will ever be perfect, that's impossible. If we had like 40 years of SC2 that would have been somewhat the same we would have a good metric to decide by, but considering how much shift and change there has been in the ~15 years lifespan of this game, it will always come down to what statistic you highlight and what metric you choose. But sadly, your way just wasn't solid at all. It was highly biased and too radical to correctly adjust for change (not that you particularly need to). But the funniest part is honestly that you think your approach is completly unbiased... Yeah, this measurement is quite faulty I think, but it brings a great point across which ironically is the one OP quoted in his post "If you torture the data long enough, it will confess to anything" At the end of the day you can use all the fancy stats and data you want, but it doesn't make your ranking any less subjective than someone's guts feeling, whether it's premobeats way overvaluing things like winrates or aligulac rating to prob Serral or the thing ejozl did here. And we all know the one true Goat ranking is this one: 1. Serral 2. Maru (extremely close to Serral) 3. Inno 4. Rogue 5. Life 6. Zest 7. sOs 8. herO 9. Dark 10. Stats
I think that quote was put there unironically to highlight the sarcastic over-the-top nature of the article.
But why do you mischaracterize what was happening? I wrote in the article that I did not come up with this weighting. I even posted the exact quote that I inserted into ChatGTP so that everyone can check it. Since then, I admitted several times that the weighting is off and even proposed a different one, where Aligulac and match win rates only make up 5% each. Also I never propped Serral lol. He gets penalized the most by subjective decisions I had to make (including non-prime years in tournament win percentage and leave out the inflation-bonus in the match win rate).
I remember being asked in the comments of my first article, if I would post the results, if changing the methodology of the original article would mean that Serral would not come out on top in any of the metrics. Well, as it is obvious now, I have no issue accepting that Serral is not the best in tournament win percentage. But it seems like people are trying to critique my result without actually addressing any core methodological feature (except for weighting, which was explained weeks ago).
What I would find more troubling than obvious weighting issues is a GOAT ranking that presents itself as statistical but in the end is a post-hoc feeling based scripture.
And of course you can make a ranking that is a lot more objective than someone's gut feeling. Not being in alignment with every subjective feeling for era-difference or tournament distinctions does not mean that going as close as possible in a statistical evaluation is the same as a random StarCraft II player that thinks this or that player is the GOAT.
We can very clearly see from the statistical evaluation that Serral is way ahead of Maru and Rogue in the time-frame that they played together. We can see that Life was ahead of Rain. We can also see what everyone subjectively thought for a long time: That Rogue has pretty mixed results.
Plus, subjectivity can be put into a bare minimum.
1. We need a point/ranking system for the tournaments (similar to what Miz and I did)
2. We need some kind of era-multiplier for different time brackets in terms of how hard it was to win (This can also be calculated more or less ok-ish as well) OR come to the conclusion that all time periods were the same, which I doubt is realistic or correct
3. Decide which metrics are the most important and then add a weighting (the community here and on reddit already agree that tournament score and tournament win participation are the most important
These are all the subjective things one needs and the community will probably come to a close proximity. But, even in the case that we have heavy disagreements here: When I revise my method one more time, I will try to make it in a way that people can tell me their desired ratios and I can post the result based on these suggestions. When all my excels are interlinked, I can easily take community suggestions for these three and we can look at how hard we would need to torture the data to make this or that player come out on top.
|
On July 13 2025 21:29 PremoBeats wrote:Show nested quote +On July 13 2025 05:44 Charoisaur wrote:On July 12 2025 06:50 Balnazza wrote: but at the same time, it doesn't wank itself into a corner by bias after bias This is literally what you did though? Just to take the WCS eu/am again, it completly fulfills your criteria and yet you disregard it. Yes, the qualifier was open to ALL koreans, they just had to move to Europe or America. That is essentially the same requirement GSL had, you had to live in Korea for a while. The biggest reason not more koreans did it was most likely Proleague, because you couldn't really play Proleague when you played WCS. But that isn't an excluding factor. This was pre-regionlock, when koreans were actually forced to be residents of Europe or America. But the whole criticism before that was how lackluster some koreans were in "their" new region, how little they contributed to the regional scenes. If you want to take prizemoney, just take prizemoney. And yes, then you will have to live with the fact that Clem (by winning EWC), Reynor (by winning Gamers8) and even Maru to a degree (by winning WESG) have a bit of inflated stats, but that is the hand you have been dealt with if you pick prizemoney as the defining statistic. But the second you introduced your "fine-tuning", you made yourself vulnerable...and as you notice, a lot of people (if not all of them) don't agree with your fine-tuning, not even the ones that agree with the outcome. Because it is such a biased and random bunch of multipliers. Just as an example I will point out again the HSC I mentioned earlier, that had barely anyone of notice in attendance but gave out more points than Katowice the next year, just because both tournaments were in different "eras"...divided by two months. damn did the scene drop in two months! And don't get me started on the highly subjective "imbalance"-multiplier... No GOAT list will ever be perfect, that's impossible. If we had like 40 years of SC2 that would have been somewhat the same we would have a good metric to decide by, but considering how much shift and change there has been in the ~15 years lifespan of this game, it will always come down to what statistic you highlight and what metric you choose. But sadly, your way just wasn't solid at all. It was highly biased and too radical to correctly adjust for change (not that you particularly need to). But the funniest part is honestly that you think your approach is completly unbiased... Yeah, this measurement is quite faulty I think, but it brings a great point across which ironically is the one OP quoted in his post "If you torture the data long enough, it will confess to anything" At the end of the day you can use all the fancy stats and data you want, but it doesn't make your ranking any less subjective than someone's guts feeling, whether it's premobeats way overvaluing things like winrates or aligulac rating to prob Serral or the thing ejozl did here. And we all know the one true Goat ranking is this one: 1. Serral 2. Maru (extremely close to Serral) 3. Inno 4. Rogue 5. Life 6. Zest 7. sOs 8. herO 9. Dark 10. Stats I think that quote was put there unironically to highlight the sarcastic over-the-top nature of the article. But why do you mischaracterize what was happening? I wrote in the article that I did not come up with this weighting. I even posted the exact quote that I inserted into ChatGTP so that everyone can check it. Since then, I admitted several times that the weighting is off and even proposed a different one, where Aligulac and match win rates only make up 5% each. Also I never propped Serral lol. He gets penalized the most by subjective decisions I had to make (including non-prime years in tournament win percentage and leave out the inflation-bonus in the match win rate). I remember being asked in the comments of my first article, if I would post the results, if changing the methodology of the original article would mean that Serral would not come out on top in any of the metrics. Well, as it is obvious now, I have no issue accepting that Serral is not the best in tournament win percentage. But it seems like people are trying to critique my result without actually addressing any core methodological feature (except for weighting, which was explained weeks ago). What I would find more troubling than obvious weighting issues is a GOAT ranking that presents itself as statistical but in the end is a post-hoc feeling based scripture. And of course you can make a ranking that is a lot more objective than someone's gut feeling. Not being in alignment with every subjective feeling for era-difference or tournament distinctions does not mean that going as close as possible in a statistical evaluation is the same as a random StarCraft II player that thinks this or that player is the GOAT. We can very clearly see from the statistical evaluation that Serral is way ahead of Maru and Rogue in the time-frame that they played together. We can see that Life was ahead of Rain. We can also see what everyone subjectively thought for a long time: That Rogue has pretty mixed results. Plus, subjectivity can be put into a bare minimum. 1. We need a point/ranking system for the tournaments (similar to what Miz and I did) 2. We need some kind of era-multiplier for different time brackets in terms of how hard it was to win (This can also be calculated more or less ok-ish as well) OR come to the conclusion that all time periods were the same, which I doubt is realistic or correct 3. Decide which metrics are the most important and then add a weighting (the community here and on reddit already agree that tournament score and tournament win participation are the most important These are all the subjective things one needs and the community will probably come to a close proximity. But, even in the case that we have heavy disagreements here: When I revise my method one more time, I will try to make it in a way that people can tell me their desired ratios and I can post the result based on these suggestions. When all my excels are interlinked, I can easily take community suggestions for these three and we can look at how hard we would need to torture the data to make this or that player come out on top.
You do not penalize Serral, though I think we all understand rhetorically why you say you do. You give him bonuses. A lot of them. And you only admit it when people call out the more obvious and ridiculous ways in which you do it like your 20% Aligulac weighting. The basic logical error you make is that you are calling all your criteria modifiers "penalties" to Serral but whether they are penalties or bonuses entirely depends on baselines versus expectations in your modifiers.
Take your tournament win % criteria. Again you call your 20% (or whatever it was) modifier a "penalty" to Serral. But the question is whether that is too high or too low or just right given that Serral competed exclusively in tournaments that are significantly easier to win than KILs, which have dozens and sometimes hundreds of additional competitors, harder groups stages, more group stages, etc. If it's just right, there's no penalty or bonus to Serral, if it's too high, it's a penalty to Serral, and if it's too low, it's a bonus to Serral. I simply do not agree that the 20% discount reflects the comparative odds of winning a 2020 Dreamhack Masters versus a 2013 GSL or SSL. It's probably 5X easier to win the former than the latter imho.
In this specific example, though, it doesn't even matter because it's actually a bonus to Serral irrespective of whether the 20% modifier is right or wrong as he is the only player on your GOAT list that does not have his tournament win % diluted by participation in the signiifcantly-harder-to-win KILs.
This is only one example of how you claim to be penalizing Serral when you're actually giving him bonuses in your model. This kind of logical error permeates your model.
At first I was giving you the benefit of the doubt that you did not understand some of these concepts in predictive modeling, but it's pretty clear that you do and are just hoping that others do not understand so you can get away with saying over and over again that you are penalizing Serral. Your entire goal is to say that no matter how much you penalize Serral, he is the GOAT, and because you're so focused on this, it's producing many errors and requiring you to go to great lengths to obscure what it is you are doing.
The simple fact is that your model giving Serral 2X the GOAT points of the next highest GOAT and 5X the GOAT points of Rogue is on its face not credible. You try to get around this by hiding behind your methodology, but in just a very cursory review of your methodology like I've done here reveals huge errors, almost all of which are designed to give Serral bonus GOAT points.
Wouldn't it be better at this point just to admin you've created a Serral = GOAT calculator? It's totally legitimate to call Serral the GOAT--he's probably most people's GOAT, even most players probably think he's the GOAT. You created a highly subjective model to crown Serral the GOAT. It's no different than this GOAT list, just different subjective criteria.
|
On July 15 2025 03:59 rwala wrote:Show nested quote +On July 13 2025 21:29 PremoBeats wrote:On July 13 2025 05:44 Charoisaur wrote:On July 12 2025 06:50 Balnazza wrote: but at the same time, it doesn't wank itself into a corner by bias after bias This is literally what you did though? Just to take the WCS eu/am again, it completly fulfills your criteria and yet you disregard it. Yes, the qualifier was open to ALL koreans, they just had to move to Europe or America. That is essentially the same requirement GSL had, you had to live in Korea for a while. The biggest reason not more koreans did it was most likely Proleague, because you couldn't really play Proleague when you played WCS. But that isn't an excluding factor. This was pre-regionlock, when koreans were actually forced to be residents of Europe or America. But the whole criticism before that was how lackluster some koreans were in "their" new region, how little they contributed to the regional scenes. If you want to take prizemoney, just take prizemoney. And yes, then you will have to live with the fact that Clem (by winning EWC), Reynor (by winning Gamers8) and even Maru to a degree (by winning WESG) have a bit of inflated stats, but that is the hand you have been dealt with if you pick prizemoney as the defining statistic. But the second you introduced your "fine-tuning", you made yourself vulnerable...and as you notice, a lot of people (if not all of them) don't agree with your fine-tuning, not even the ones that agree with the outcome. Because it is such a biased and random bunch of multipliers. Just as an example I will point out again the HSC I mentioned earlier, that had barely anyone of notice in attendance but gave out more points than Katowice the next year, just because both tournaments were in different "eras"...divided by two months. damn did the scene drop in two months! And don't get me started on the highly subjective "imbalance"-multiplier... No GOAT list will ever be perfect, that's impossible. If we had like 40 years of SC2 that would have been somewhat the same we would have a good metric to decide by, but considering how much shift and change there has been in the ~15 years lifespan of this game, it will always come down to what statistic you highlight and what metric you choose. But sadly, your way just wasn't solid at all. It was highly biased and too radical to correctly adjust for change (not that you particularly need to). But the funniest part is honestly that you think your approach is completly unbiased... Yeah, this measurement is quite faulty I think, but it brings a great point across which ironically is the one OP quoted in his post "If you torture the data long enough, it will confess to anything" At the end of the day you can use all the fancy stats and data you want, but it doesn't make your ranking any less subjective than someone's guts feeling, whether it's premobeats way overvaluing things like winrates or aligulac rating to prob Serral or the thing ejozl did here. And we all know the one true Goat ranking is this one: 1. Serral 2. Maru (extremely close to Serral) 3. Inno 4. Rogue 5. Life 6. Zest 7. sOs 8. herO 9. Dark 10. Stats I think that quote was put there unironically to highlight the sarcastic over-the-top nature of the article. But why do you mischaracterize what was happening? I wrote in the article that I did not come up with this weighting. I even posted the exact quote that I inserted into ChatGTP so that everyone can check it. Since then, I admitted several times that the weighting is off and even proposed a different one, where Aligulac and match win rates only make up 5% each. Also I never propped Serral lol. He gets penalized the most by subjective decisions I had to make (including non-prime years in tournament win percentage and leave out the inflation-bonus in the match win rate). I remember being asked in the comments of my first article, if I would post the results, if changing the methodology of the original article would mean that Serral would not come out on top in any of the metrics. Well, as it is obvious now, I have no issue accepting that Serral is not the best in tournament win percentage. But it seems like people are trying to critique my result without actually addressing any core methodological feature (except for weighting, which was explained weeks ago). What I would find more troubling than obvious weighting issues is a GOAT ranking that presents itself as statistical but in the end is a post-hoc feeling based scripture. And of course you can make a ranking that is a lot more objective than someone's gut feeling. Not being in alignment with every subjective feeling for era-difference or tournament distinctions does not mean that going as close as possible in a statistical evaluation is the same as a random StarCraft II player that thinks this or that player is the GOAT. We can very clearly see from the statistical evaluation that Serral is way ahead of Maru and Rogue in the time-frame that they played together. We can see that Life was ahead of Rain. We can also see what everyone subjectively thought for a long time: That Rogue has pretty mixed results. Plus, subjectivity can be put into a bare minimum. 1. We need a point/ranking system for the tournaments (similar to what Miz and I did) 2. We need some kind of era-multiplier for different time brackets in terms of how hard it was to win (This can also be calculated more or less ok-ish as well) OR come to the conclusion that all time periods were the same, which I doubt is realistic or correct 3. Decide which metrics are the most important and then add a weighting (the community here and on reddit already agree that tournament score and tournament win participation are the most important These are all the subjective things one needs and the community will probably come to a close proximity. But, even in the case that we have heavy disagreements here: When I revise my method one more time, I will try to make it in a way that people can tell me their desired ratios and I can post the result based on these suggestions. When all my excels are interlinked, I can easily take community suggestions for these three and we can look at how hard we would need to torture the data to make this or that player come out on top. You do not penalize Serral, though I think we all understand rhetorically why you say you do. You give him bonuses. A lot of them. And you only admit it when people call out the more obvious and ridiculous ways in which you do it like your 20% Aligulac weighting. The basic logical error you make is that you are calling all your criteria modifiers "penalties" to Serral but whether they are penalties or bonuses entirely depends on baselines versus expectations in your modifiers. Take your tournament win % criteria. Again you call your 20% (or whatever it was) modifier a "penalty" to Serral. But the question is whether that is too high or too low or just right given that Serral competed exclusively in tournaments that are significantly easier to win than KILs, which have dozens and sometimes hundreds of additional competitors, harder groups stages, more group stages, etc. If it's just right, there's no penalty or bonus to Serral, if it's too high, it's a penalty to Serral, and if it's too low, it's a bonus to Serral. I simply do not agree that the 20% discount reflects the comparative odds of winning a 2020 Dreamhack Masters versus a 2013 GSL or SSL. It's probably 5X easier to win the former than the latter imho. In this specific example, though, it doesn't even matter because it's actually a bonus to Serral irrespective of whether the 20% modifier is right or wrong as he is the only player on your GOAT list that does not have his tournament win % diluted by participation in the signiifcantly-harder-to-win KILs. This is only one example of how you claim to be penalizing Serral when you're actually giving him bonuses in your model. This kind of logical error permeates your model. At first I was giving you the benefit of the doubt that you did not understand some of these concepts in predictive modeling, but it's pretty clear that you do and are just hoping that others do not understand so you can get away with saying over and over again that you are penalizing Serral. Your entire goal is to say that no matter how much you penalize Serral, he is the GOAT, and because you're so focused on this, it's producing many errors and requiring you to go to great lengths to obscure what it is you are doing. The simple fact is that your model giving Serral 2X the GOAT points of the next highest GOAT and 5X the GOAT points of Rogue is on its face not credible. You try to get around this by hiding behind your methodology, but in just a very cursory review of your methodology like I've done here reveals huge errors, almost all of which are designed to give Serral bonus GOAT points. Wouldn't it be better at this point just to admin you've created a Serral = GOAT calculator? It's totally legitimate to call Serral the GOAT--he's probably most people's GOAT, even most players probably think he's the GOAT. You created a highly subjective model to crown Serral the GOAT. It's no different than this GOAT list, just different subjective criteria.
Well said
|
On July 15 2025 17:32 Charoisaur wrote:Show nested quote +On July 15 2025 03:59 rwala wrote:On July 13 2025 21:29 PremoBeats wrote:On July 13 2025 05:44 Charoisaur wrote:On July 12 2025 06:50 Balnazza wrote: but at the same time, it doesn't wank itself into a corner by bias after bias This is literally what you did though? Just to take the WCS eu/am again, it completly fulfills your criteria and yet you disregard it. Yes, the qualifier was open to ALL koreans, they just had to move to Europe or America. That is essentially the same requirement GSL had, you had to live in Korea for a while. The biggest reason not more koreans did it was most likely Proleague, because you couldn't really play Proleague when you played WCS. But that isn't an excluding factor. This was pre-regionlock, when koreans were actually forced to be residents of Europe or America. But the whole criticism before that was how lackluster some koreans were in "their" new region, how little they contributed to the regional scenes. If you want to take prizemoney, just take prizemoney. And yes, then you will have to live with the fact that Clem (by winning EWC), Reynor (by winning Gamers8) and even Maru to a degree (by winning WESG) have a bit of inflated stats, but that is the hand you have been dealt with if you pick prizemoney as the defining statistic. But the second you introduced your "fine-tuning", you made yourself vulnerable...and as you notice, a lot of people (if not all of them) don't agree with your fine-tuning, not even the ones that agree with the outcome. Because it is such a biased and random bunch of multipliers. Just as an example I will point out again the HSC I mentioned earlier, that had barely anyone of notice in attendance but gave out more points than Katowice the next year, just because both tournaments were in different "eras"...divided by two months. damn did the scene drop in two months! And don't get me started on the highly subjective "imbalance"-multiplier... No GOAT list will ever be perfect, that's impossible. If we had like 40 years of SC2 that would have been somewhat the same we would have a good metric to decide by, but considering how much shift and change there has been in the ~15 years lifespan of this game, it will always come down to what statistic you highlight and what metric you choose. But sadly, your way just wasn't solid at all. It was highly biased and too radical to correctly adjust for change (not that you particularly need to). But the funniest part is honestly that you think your approach is completly unbiased... Yeah, this measurement is quite faulty I think, but it brings a great point across which ironically is the one OP quoted in his post "If you torture the data long enough, it will confess to anything" At the end of the day you can use all the fancy stats and data you want, but it doesn't make your ranking any less subjective than someone's guts feeling, whether it's premobeats way overvaluing things like winrates or aligulac rating to prob Serral or the thing ejozl did here. And we all know the one true Goat ranking is this one: 1. Serral 2. Maru (extremely close to Serral) 3. Inno 4. Rogue 5. Life 6. Zest 7. sOs 8. herO 9. Dark 10. Stats I think that quote was put there unironically to highlight the sarcastic over-the-top nature of the article. But why do you mischaracterize what was happening? I wrote in the article that I did not come up with this weighting. I even posted the exact quote that I inserted into ChatGTP so that everyone can check it. Since then, I admitted several times that the weighting is off and even proposed a different one, where Aligulac and match win rates only make up 5% each. Also I never propped Serral lol. He gets penalized the most by subjective decisions I had to make (including non-prime years in tournament win percentage and leave out the inflation-bonus in the match win rate). I remember being asked in the comments of my first article, if I would post the results, if changing the methodology of the original article would mean that Serral would not come out on top in any of the metrics. Well, as it is obvious now, I have no issue accepting that Serral is not the best in tournament win percentage. But it seems like people are trying to critique my result without actually addressing any core methodological feature (except for weighting, which was explained weeks ago). What I would find more troubling than obvious weighting issues is a GOAT ranking that presents itself as statistical but in the end is a post-hoc feeling based scripture. And of course you can make a ranking that is a lot more objective than someone's gut feeling. Not being in alignment with every subjective feeling for era-difference or tournament distinctions does not mean that going as close as possible in a statistical evaluation is the same as a random StarCraft II player that thinks this or that player is the GOAT. We can very clearly see from the statistical evaluation that Serral is way ahead of Maru and Rogue in the time-frame that they played together. We can see that Life was ahead of Rain. We can also see what everyone subjectively thought for a long time: That Rogue has pretty mixed results. Plus, subjectivity can be put into a bare minimum. 1. We need a point/ranking system for the tournaments (similar to what Miz and I did) 2. We need some kind of era-multiplier for different time brackets in terms of how hard it was to win (This can also be calculated more or less ok-ish as well) OR come to the conclusion that all time periods were the same, which I doubt is realistic or correct 3. Decide which metrics are the most important and then add a weighting (the community here and on reddit already agree that tournament score and tournament win participation are the most important These are all the subjective things one needs and the community will probably come to a close proximity. But, even in the case that we have heavy disagreements here: When I revise my method one more time, I will try to make it in a way that people can tell me their desired ratios and I can post the result based on these suggestions. When all my excels are interlinked, I can easily take community suggestions for these three and we can look at how hard we would need to torture the data to make this or that player come out on top. You do not penalize Serral, though I think we all understand rhetorically why you say you do. You give him bonuses. A lot of them. And you only admit it when people call out the more obvious and ridiculous ways in which you do it like your 20% Aligulac weighting. The basic logical error you make is that you are calling all your criteria modifiers "penalties" to Serral but whether they are penalties or bonuses entirely depends on baselines versus expectations in your modifiers. Take your tournament win % criteria. Again you call your 20% (or whatever it was) modifier a "penalty" to Serral. But the question is whether that is too high or too low or just right given that Serral competed exclusively in tournaments that are significantly easier to win than KILs, which have dozens and sometimes hundreds of additional competitors, harder groups stages, more group stages, etc. If it's just right, there's no penalty or bonus to Serral, if it's too high, it's a penalty to Serral, and if it's too low, it's a bonus to Serral. I simply do not agree that the 20% discount reflects the comparative odds of winning a 2020 Dreamhack Masters versus a 2013 GSL or SSL. It's probably 5X easier to win the former than the latter imho. In this specific example, though, it doesn't even matter because it's actually a bonus to Serral irrespective of whether the 20% modifier is right or wrong as he is the only player on your GOAT list that does not have his tournament win % diluted by participation in the signiifcantly-harder-to-win KILs. This is only one example of how you claim to be penalizing Serral when you're actually giving him bonuses in your model. This kind of logical error permeates your model. At first I was giving you the benefit of the doubt that you did not understand some of these concepts in predictive modeling, but it's pretty clear that you do and are just hoping that others do not understand so you can get away with saying over and over again that you are penalizing Serral. Your entire goal is to say that no matter how much you penalize Serral, he is the GOAT, and because you're so focused on this, it's producing many errors and requiring you to go to great lengths to obscure what it is you are doing. The simple fact is that your model giving Serral 2X the GOAT points of the next highest GOAT and 5X the GOAT points of Rogue is on its face not credible. You try to get around this by hiding behind your methodology, but in just a very cursory review of your methodology like I've done here reveals huge errors, almost all of which are designed to give Serral bonus GOAT points. Wouldn't it be better at this point just to admin you've created a Serral = GOAT calculator? It's totally legitimate to call Serral the GOAT--he's probably most people's GOAT, even most players probably think he's the GOAT. You created a highly subjective model to crown Serral the GOAT. It's no different than this GOAT list, just different subjective criteria. Well said
rwala's reply consists mostly of straw men, false comparisons and already addressed arguments... what exactly do you think is well said about it?
On July 15 2025 03:59 rwala wrote:Show nested quote +On July 13 2025 21:29 PremoBeats wrote:On July 13 2025 05:44 Charoisaur wrote:On July 12 2025 06:50 Balnazza wrote: but at the same time, it doesn't wank itself into a corner by bias after bias This is literally what you did though? Just to take the WCS eu/am again, it completly fulfills your criteria and yet you disregard it. Yes, the qualifier was open to ALL koreans, they just had to move to Europe or America. That is essentially the same requirement GSL had, you had to live in Korea for a while. The biggest reason not more koreans did it was most likely Proleague, because you couldn't really play Proleague when you played WCS. But that isn't an excluding factor. This was pre-regionlock, when koreans were actually forced to be residents of Europe or America. But the whole criticism before that was how lackluster some koreans were in "their" new region, how little they contributed to the regional scenes. If you want to take prizemoney, just take prizemoney. And yes, then you will have to live with the fact that Clem (by winning EWC), Reynor (by winning Gamers8) and even Maru to a degree (by winning WESG) have a bit of inflated stats, but that is the hand you have been dealt with if you pick prizemoney as the defining statistic. But the second you introduced your "fine-tuning", you made yourself vulnerable...and as you notice, a lot of people (if not all of them) don't agree with your fine-tuning, not even the ones that agree with the outcome. Because it is such a biased and random bunch of multipliers. Just as an example I will point out again the HSC I mentioned earlier, that had barely anyone of notice in attendance but gave out more points than Katowice the next year, just because both tournaments were in different "eras"...divided by two months. damn did the scene drop in two months! And don't get me started on the highly subjective "imbalance"-multiplier... No GOAT list will ever be perfect, that's impossible. If we had like 40 years of SC2 that would have been somewhat the same we would have a good metric to decide by, but considering how much shift and change there has been in the ~15 years lifespan of this game, it will always come down to what statistic you highlight and what metric you choose. But sadly, your way just wasn't solid at all. It was highly biased and too radical to correctly adjust for change (not that you particularly need to). But the funniest part is honestly that you think your approach is completly unbiased... Yeah, this measurement is quite faulty I think, but it brings a great point across which ironically is the one OP quoted in his post "If you torture the data long enough, it will confess to anything" At the end of the day you can use all the fancy stats and data you want, but it doesn't make your ranking any less subjective than someone's guts feeling, whether it's premobeats way overvaluing things like winrates or aligulac rating to prob Serral or the thing ejozl did here. And we all know the one true Goat ranking is this one: 1. Serral 2. Maru (extremely close to Serral) 3. Inno 4. Rogue 5. Life 6. Zest 7. sOs 8. herO 9. Dark 10. Stats I think that quote was put there unironically to highlight the sarcastic over-the-top nature of the article. But why do you mischaracterize what was happening? I wrote in the article that I did not come up with this weighting. I even posted the exact quote that I inserted into ChatGTP so that everyone can check it. Since then, I admitted several times that the weighting is off and even proposed a different one, where Aligulac and match win rates only make up 5% each. Also I never propped Serral lol. He gets penalized the most by subjective decisions I had to make (including non-prime years in tournament win percentage and leave out the inflation-bonus in the match win rate). I remember being asked in the comments of my first article, if I would post the results, if changing the methodology of the original article would mean that Serral would not come out on top in any of the metrics. Well, as it is obvious now, I have no issue accepting that Serral is not the best in tournament win percentage. But it seems like people are trying to critique my result without actually addressing any core methodological feature (except for weighting, which was explained weeks ago). What I would find more troubling than obvious weighting issues is a GOAT ranking that presents itself as statistical but in the end is a post-hoc feeling based scripture. And of course you can make a ranking that is a lot more objective than someone's gut feeling. Not being in alignment with every subjective feeling for era-difference or tournament distinctions does not mean that going as close as possible in a statistical evaluation is the same as a random StarCraft II player that thinks this or that player is the GOAT. We can very clearly see from the statistical evaluation that Serral is way ahead of Maru and Rogue in the time-frame that they played together. We can see that Life was ahead of Rain. We can also see what everyone subjectively thought for a long time: That Rogue has pretty mixed results. Plus, subjectivity can be put into a bare minimum. 1. We need a point/ranking system for the tournaments (similar to what Miz and I did) 2. We need some kind of era-multiplier for different time brackets in terms of how hard it was to win (This can also be calculated more or less ok-ish as well) OR come to the conclusion that all time periods were the same, which I doubt is realistic or correct 3. Decide which metrics are the most important and then add a weighting (the community here and on reddit already agree that tournament score and tournament win participation are the most important These are all the subjective things one needs and the community will probably come to a close proximity. But, even in the case that we have heavy disagreements here: When I revise my method one more time, I will try to make it in a way that people can tell me their desired ratios and I can post the result based on these suggestions. When all my excels are interlinked, I can easily take community suggestions for these three and we can look at how hard we would need to torture the data to make this or that player come out on top. You do not penalize Serral, though I think we all understand rhetorically why you say you do. You give him bonuses. A lot of them. And you only admit it when people call out the more obvious and ridiculous ways in which you do it like your 20% Aligulac weighting. The basic logical error you make is that you are calling all your criteria modifiers "penalties" to Serral but whether they are penalties or bonuses entirely depends on baselines versus expectations in your modifiers. Take your tournament win % criteria. Again you call your 20% (or whatever it was) modifier a "penalty" to Serral. But the question is whether that is too high or too low or just right given that Serral competed exclusively in tournaments that are significantly easier to win than KILs, which have dozens and sometimes hundreds of additional competitors, harder groups stages, more group stages, etc. If it's just right, there's no penalty or bonus to Serral, if it's too high, it's a penalty to Serral, and if it's too low, it's a bonus to Serral. I simply do not agree that the 20% discount reflects the comparative odds of winning a 2020 Dreamhack Masters versus a 2013 GSL or SSL. It's probably 5X easier to win the former than the latter imho. In this specific example, though, it doesn't even matter because it's actually a bonus to Serral irrespective of whether the 20% modifier is right or wrong as he is the only player on your GOAT list that does not have his tournament win % diluted by participation in the signiifcantly-harder-to-win KILs. This is only one example of how you claim to be penalizing Serral when you're actually giving him bonuses in your model. This kind of logical error permeates your model. At first I was giving you the benefit of the doubt that you did not understand some of these concepts in predictive modeling, but it's pretty clear that you do and are just hoping that others do not understand so you can get away with saying over and over again that you are penalizing Serral. Your entire goal is to say that no matter how much you penalize Serral, he is the GOAT, and because you're so focused on this, it's producing many errors and requiring you to go to great lengths to obscure what it is you are doing. The simple fact is that your model giving Serral 2X the GOAT points of the next highest GOAT and 5X the GOAT points of Rogue is on its face not credible. You try to get around this by hiding behind your methodology, but in just a very cursory review of your methodology like I've done here reveals huge errors, almost all of which are designed to give Serral bonus GOAT points. Wouldn't it be better at this point just to admin you've created a Serral = GOAT calculator? It's totally legitimate to call Serral the GOAT--he's probably most people's GOAT, even most players probably think he's the GOAT. You created a highly subjective model to crown Serral the GOAT. It's no different than this GOAT list, just different subjective criteria.
It is not rhetorical to say that I penalized Serral, because I did:
1a. Do you deny that the Koreans played more lower tier players in comparison to Serral? Or do you deny that I wrote that I did not include this correction, which is detrimental to Serral?
1b. Do you deny that including non-prime years, which I did, penalizes/hurts Serral the most according to the data?
2. In the article - as well as multiple times in the other thread (and even in the very post you commented on and directly to you in the other thread!!) I explained that this weighting was proposed by ChatGTP.
I still don't understand why you keep pounding on that, tbh. As WombaT wrote out in the other thread, the weighting is entirely superfluous. The most important revelation was that most players have mixed results, while Serral is dominating them all.
You make it seem like the weighting was there to push Serral, which is not the case. You have no proof for this accusation. On the contrary, I posted the exact text that I used to let ChatGTP decide on the weighting, to let you guys double check. I only did so because I knew based on the results, that the weighting does not change in the first place and I couldn’t decide on a specific weighting, which I explained ad nauseam in the article as well as in the comments. First place would only change with an insane weighting-setup for efficiency… then Life (who most people disregard anyway) would be on top, with Serral being 2nd.
Arguing with the weighting result is utterly pointless… As this is your (and other critics) biggest point which you all come back to every time, I doubt that is much more substance to gain from you guys.
3a. In the tournament win %, similar to most metrics, I discounted all region-locked tournaments. This is a penalty to Serral, although a needed one to deny his easier tournament wins. But still it is a penalty; especially to deny them entirely, as Korean pros had rather easy tournament wins as well.
3b. You might disagree that the bonus is not enough according to your subjective feeling.
My reasoning for using 20%: I explained in detail why a larger player pool, especially in mostly tier 2, 3 and 4 players, would not decrease a tournament win percentage that much.
- doubling the player pool would most likely mean just one extra round. If that round was a group stage, it would naturally mean that you’d have to play more lower tier player statistically, boosting your match win rate
- I reasoned that this fact means that average place would need a higher era-multiplier than tournament win percentage
- Now looking at Serral’s average win rate of around 82% in his years from 2018-2024 subtracted with 2% (some lower match win for era but also considering that his win rates are deflated versus the Koreans because of 1a, also taking into account that earlier group stages are easier to win than knockout brackets), these more rounds would mean a win probability of roughly 80%. As an example: going from 4 rounds to 5 rounds would statistically mean a win probability dropping from 40% to 33%, meaning a loss of 17,5% or a boost of 21,2%.
- Now, comparing DH 2020 to a GSL or SSL 2013: Not all tournaments that other players have won were a GSL though. What about the 2013 Hot6ix Cup or ASUS ROG NorthCon? And didn’t Serral win uberly stacked tournaments as well?
You are committing a false equivalence/improper comparison fallacy here. In essence, you are right that one can’t compare a GSL to a DH. But that is also not what I did. There were not only GSLs in 2013 (as a matter of fact, there was only one) and Serral won more stacked tournaments as well. It also borders on a strawman, as my approach was not designed to precisely reflect the odds between only one of the hardest 2013 tournaments vs a specific 2020 event, but rather roughly corrects across periods.
- Do you further realize that arguing like this from an era/period-perspective also argues against Rogue’s and Maru’s case?
- What adjustment would you find appropriate?
Overall it is rather the modern Koreans that benefitted in this metric, as Serral, Clem and mostly Reynor were missing in the GSLs that they won.
Of course Serral wouldn’t have the same tournament win percentage had he been a full-time pro at his peak playing in the prime era. But the win percentages of INno, Rain, etc. would drop as well, had Mvp, Serral and Clem played there. I explained this already in the article.
“Again you call your 20% (or whatever it was) modifier a "penalty" to Serral”
Can you give the exact quote?
This would be much more fruitful if you didn’t try to rely on notions like my method was not statistical, keep using arguments that have been addressed or straight out deny that Serral was penalized on several occasions, which he clearly was.
Talk about the points ratio in the tournaments or the era-multiplier or your preferred weighting. Or - as you by your own account don’t care for statistical GOAT debates anyway - simply stay out of them. But making unsubstantiated accusations like me making a Serral GOAT-calculator or repeating the same already addressed notions is absolutely ridiculous.
|
On July 15 2025 19:31 PremoBeats wrote:Show nested quote +On July 15 2025 17:32 Charoisaur wrote:On July 15 2025 03:59 rwala wrote:On July 13 2025 21:29 PremoBeats wrote:On July 13 2025 05:44 Charoisaur wrote:On July 12 2025 06:50 Balnazza wrote: but at the same time, it doesn't wank itself into a corner by bias after bias This is literally what you did though? Just to take the WCS eu/am again, it completly fulfills your criteria and yet you disregard it. Yes, the qualifier was open to ALL koreans, they just had to move to Europe or America. That is essentially the same requirement GSL had, you had to live in Korea for a while. The biggest reason not more koreans did it was most likely Proleague, because you couldn't really play Proleague when you played WCS. But that isn't an excluding factor. This was pre-regionlock, when koreans were actually forced to be residents of Europe or America. But the whole criticism before that was how lackluster some koreans were in "their" new region, how little they contributed to the regional scenes. If you want to take prizemoney, just take prizemoney. And yes, then you will have to live with the fact that Clem (by winning EWC), Reynor (by winning Gamers8) and even Maru to a degree (by winning WESG) have a bit of inflated stats, but that is the hand you have been dealt with if you pick prizemoney as the defining statistic. But the second you introduced your "fine-tuning", you made yourself vulnerable...and as you notice, a lot of people (if not all of them) don't agree with your fine-tuning, not even the ones that agree with the outcome. Because it is such a biased and random bunch of multipliers. Just as an example I will point out again the HSC I mentioned earlier, that had barely anyone of notice in attendance but gave out more points than Katowice the next year, just because both tournaments were in different "eras"...divided by two months. damn did the scene drop in two months! And don't get me started on the highly subjective "imbalance"-multiplier... No GOAT list will ever be perfect, that's impossible. If we had like 40 years of SC2 that would have been somewhat the same we would have a good metric to decide by, but considering how much shift and change there has been in the ~15 years lifespan of this game, it will always come down to what statistic you highlight and what metric you choose. But sadly, your way just wasn't solid at all. It was highly biased and too radical to correctly adjust for change (not that you particularly need to). But the funniest part is honestly that you think your approach is completly unbiased... Yeah, this measurement is quite faulty I think, but it brings a great point across which ironically is the one OP quoted in his post "If you torture the data long enough, it will confess to anything" At the end of the day you can use all the fancy stats and data you want, but it doesn't make your ranking any less subjective than someone's guts feeling, whether it's premobeats way overvaluing things like winrates or aligulac rating to prob Serral or the thing ejozl did here. And we all know the one true Goat ranking is this one: 1. Serral 2. Maru (extremely close to Serral) 3. Inno 4. Rogue 5. Life 6. Zest 7. sOs 8. herO 9. Dark 10. Stats I think that quote was put there unironically to highlight the sarcastic over-the-top nature of the article. But why do you mischaracterize what was happening? I wrote in the article that I did not come up with this weighting. I even posted the exact quote that I inserted into ChatGTP so that everyone can check it. Since then, I admitted several times that the weighting is off and even proposed a different one, where Aligulac and match win rates only make up 5% each. Also I never propped Serral lol. He gets penalized the most by subjective decisions I had to make (including non-prime years in tournament win percentage and leave out the inflation-bonus in the match win rate). I remember being asked in the comments of my first article, if I would post the results, if changing the methodology of the original article would mean that Serral would not come out on top in any of the metrics. Well, as it is obvious now, I have no issue accepting that Serral is not the best in tournament win percentage. But it seems like people are trying to critique my result without actually addressing any core methodological feature (except for weighting, which was explained weeks ago). What I would find more troubling than obvious weighting issues is a GOAT ranking that presents itself as statistical but in the end is a post-hoc feeling based scripture. And of course you can make a ranking that is a lot more objective than someone's gut feeling. Not being in alignment with every subjective feeling for era-difference or tournament distinctions does not mean that going as close as possible in a statistical evaluation is the same as a random StarCraft II player that thinks this or that player is the GOAT. We can very clearly see from the statistical evaluation that Serral is way ahead of Maru and Rogue in the time-frame that they played together. We can see that Life was ahead of Rain. We can also see what everyone subjectively thought for a long time: That Rogue has pretty mixed results. Plus, subjectivity can be put into a bare minimum. 1. We need a point/ranking system for the tournaments (similar to what Miz and I did) 2. We need some kind of era-multiplier for different time brackets in terms of how hard it was to win (This can also be calculated more or less ok-ish as well) OR come to the conclusion that all time periods were the same, which I doubt is realistic or correct 3. Decide which metrics are the most important and then add a weighting (the community here and on reddit already agree that tournament score and tournament win participation are the most important These are all the subjective things one needs and the community will probably come to a close proximity. But, even in the case that we have heavy disagreements here: When I revise my method one more time, I will try to make it in a way that people can tell me their desired ratios and I can post the result based on these suggestions. When all my excels are interlinked, I can easily take community suggestions for these three and we can look at how hard we would need to torture the data to make this or that player come out on top. You do not penalize Serral, though I think we all understand rhetorically why you say you do. You give him bonuses. A lot of them. And you only admit it when people call out the more obvious and ridiculous ways in which you do it like your 20% Aligulac weighting. The basic logical error you make is that you are calling all your criteria modifiers "penalties" to Serral but whether they are penalties or bonuses entirely depends on baselines versus expectations in your modifiers. Take your tournament win % criteria. Again you call your 20% (or whatever it was) modifier a "penalty" to Serral. But the question is whether that is too high or too low or just right given that Serral competed exclusively in tournaments that are significantly easier to win than KILs, which have dozens and sometimes hundreds of additional competitors, harder groups stages, more group stages, etc. If it's just right, there's no penalty or bonus to Serral, if it's too high, it's a penalty to Serral, and if it's too low, it's a bonus to Serral. I simply do not agree that the 20% discount reflects the comparative odds of winning a 2020 Dreamhack Masters versus a 2013 GSL or SSL. It's probably 5X easier to win the former than the latter imho. In this specific example, though, it doesn't even matter because it's actually a bonus to Serral irrespective of whether the 20% modifier is right or wrong as he is the only player on your GOAT list that does not have his tournament win % diluted by participation in the signiifcantly-harder-to-win KILs. This is only one example of how you claim to be penalizing Serral when you're actually giving him bonuses in your model. This kind of logical error permeates your model. At first I was giving you the benefit of the doubt that you did not understand some of these concepts in predictive modeling, but it's pretty clear that you do and are just hoping that others do not understand so you can get away with saying over and over again that you are penalizing Serral. Your entire goal is to say that no matter how much you penalize Serral, he is the GOAT, and because you're so focused on this, it's producing many errors and requiring you to go to great lengths to obscure what it is you are doing. The simple fact is that your model giving Serral 2X the GOAT points of the next highest GOAT and 5X the GOAT points of Rogue is on its face not credible. You try to get around this by hiding behind your methodology, but in just a very cursory review of your methodology like I've done here reveals huge errors, almost all of which are designed to give Serral bonus GOAT points. Wouldn't it be better at this point just to admin you've created a Serral = GOAT calculator? It's totally legitimate to call Serral the GOAT--he's probably most people's GOAT, even most players probably think he's the GOAT. You created a highly subjective model to crown Serral the GOAT. It's no different than this GOAT list, just different subjective criteria. Well said rwala's reply consists mostly of straw men, false comparisons and already addressed arguments... what exactly do you think is well said about it? Show nested quote +On July 15 2025 03:59 rwala wrote:On July 13 2025 21:29 PremoBeats wrote:On July 13 2025 05:44 Charoisaur wrote:On July 12 2025 06:50 Balnazza wrote: but at the same time, it doesn't wank itself into a corner by bias after bias This is literally what you did though? Just to take the WCS eu/am again, it completly fulfills your criteria and yet you disregard it. Yes, the qualifier was open to ALL koreans, they just had to move to Europe or America. That is essentially the same requirement GSL had, you had to live in Korea for a while. The biggest reason not more koreans did it was most likely Proleague, because you couldn't really play Proleague when you played WCS. But that isn't an excluding factor. This was pre-regionlock, when koreans were actually forced to be residents of Europe or America. But the whole criticism before that was how lackluster some koreans were in "their" new region, how little they contributed to the regional scenes. If you want to take prizemoney, just take prizemoney. And yes, then you will have to live with the fact that Clem (by winning EWC), Reynor (by winning Gamers8) and even Maru to a degree (by winning WESG) have a bit of inflated stats, but that is the hand you have been dealt with if you pick prizemoney as the defining statistic. But the second you introduced your "fine-tuning", you made yourself vulnerable...and as you notice, a lot of people (if not all of them) don't agree with your fine-tuning, not even the ones that agree with the outcome. Because it is such a biased and random bunch of multipliers. Just as an example I will point out again the HSC I mentioned earlier, that had barely anyone of notice in attendance but gave out more points than Katowice the next year, just because both tournaments were in different "eras"...divided by two months. damn did the scene drop in two months! And don't get me started on the highly subjective "imbalance"-multiplier... No GOAT list will ever be perfect, that's impossible. If we had like 40 years of SC2 that would have been somewhat the same we would have a good metric to decide by, but considering how much shift and change there has been in the ~15 years lifespan of this game, it will always come down to what statistic you highlight and what metric you choose. But sadly, your way just wasn't solid at all. It was highly biased and too radical to correctly adjust for change (not that you particularly need to). But the funniest part is honestly that you think your approach is completly unbiased... Yeah, this measurement is quite faulty I think, but it brings a great point across which ironically is the one OP quoted in his post "If you torture the data long enough, it will confess to anything" At the end of the day you can use all the fancy stats and data you want, but it doesn't make your ranking any less subjective than someone's guts feeling, whether it's premobeats way overvaluing things like winrates or aligulac rating to prob Serral or the thing ejozl did here. And we all know the one true Goat ranking is this one: 1. Serral 2. Maru (extremely close to Serral) 3. Inno 4. Rogue 5. Life 6. Zest 7. sOs 8. herO 9. Dark 10. Stats I think that quote was put there unironically to highlight the sarcastic over-the-top nature of the article. But why do you mischaracterize what was happening? I wrote in the article that I did not come up with this weighting. I even posted the exact quote that I inserted into ChatGTP so that everyone can check it. Since then, I admitted several times that the weighting is off and even proposed a different one, where Aligulac and match win rates only make up 5% each. Also I never propped Serral lol. He gets penalized the most by subjective decisions I had to make (including non-prime years in tournament win percentage and leave out the inflation-bonus in the match win rate). I remember being asked in the comments of my first article, if I would post the results, if changing the methodology of the original article would mean that Serral would not come out on top in any of the metrics. Well, as it is obvious now, I have no issue accepting that Serral is not the best in tournament win percentage. But it seems like people are trying to critique my result without actually addressing any core methodological feature (except for weighting, which was explained weeks ago). What I would find more troubling than obvious weighting issues is a GOAT ranking that presents itself as statistical but in the end is a post-hoc feeling based scripture. And of course you can make a ranking that is a lot more objective than someone's gut feeling. Not being in alignment with every subjective feeling for era-difference or tournament distinctions does not mean that going as close as possible in a statistical evaluation is the same as a random StarCraft II player that thinks this or that player is the GOAT. We can very clearly see from the statistical evaluation that Serral is way ahead of Maru and Rogue in the time-frame that they played together. We can see that Life was ahead of Rain. We can also see what everyone subjectively thought for a long time: That Rogue has pretty mixed results. Plus, subjectivity can be put into a bare minimum. 1. We need a point/ranking system for the tournaments (similar to what Miz and I did) 2. We need some kind of era-multiplier for different time brackets in terms of how hard it was to win (This can also be calculated more or less ok-ish as well) OR come to the conclusion that all time periods were the same, which I doubt is realistic or correct 3. Decide which metrics are the most important and then add a weighting (the community here and on reddit already agree that tournament score and tournament win participation are the most important These are all the subjective things one needs and the community will probably come to a close proximity. But, even in the case that we have heavy disagreements here: When I revise my method one more time, I will try to make it in a way that people can tell me their desired ratios and I can post the result based on these suggestions. When all my excels are interlinked, I can easily take community suggestions for these three and we can look at how hard we would need to torture the data to make this or that player come out on top. You do not penalize Serral, though I think we all understand rhetorically why you say you do. You give him bonuses. A lot of them. And you only admit it when people call out the more obvious and ridiculous ways in which you do it like your 20% Aligulac weighting. The basic logical error you make is that you are calling all your criteria modifiers "penalties" to Serral but whether they are penalties or bonuses entirely depends on baselines versus expectations in your modifiers. Take your tournament win % criteria. Again you call your 20% (or whatever it was) modifier a "penalty" to Serral. But the question is whether that is too high or too low or just right given that Serral competed exclusively in tournaments that are significantly easier to win than KILs, which have dozens and sometimes hundreds of additional competitors, harder groups stages, more group stages, etc. If it's just right, there's no penalty or bonus to Serral, if it's too high, it's a penalty to Serral, and if it's too low, it's a bonus to Serral. I simply do not agree that the 20% discount reflects the comparative odds of winning a 2020 Dreamhack Masters versus a 2013 GSL or SSL. It's probably 5X easier to win the former than the latter imho. In this specific example, though, it doesn't even matter because it's actually a bonus to Serral irrespective of whether the 20% modifier is right or wrong as he is the only player on your GOAT list that does not have his tournament win % diluted by participation in the signiifcantly-harder-to-win KILs. This is only one example of how you claim to be penalizing Serral when you're actually giving him bonuses in your model. This kind of logical error permeates your model. At first I was giving you the benefit of the doubt that you did not understand some of these concepts in predictive modeling, but it's pretty clear that you do and are just hoping that others do not understand so you can get away with saying over and over again that you are penalizing Serral. Your entire goal is to say that no matter how much you penalize Serral, he is the GOAT, and because you're so focused on this, it's producing many errors and requiring you to go to great lengths to obscure what it is you are doing. The simple fact is that your model giving Serral 2X the GOAT points of the next highest GOAT and 5X the GOAT points of Rogue is on its face not credible. You try to get around this by hiding behind your methodology, but in just a very cursory review of your methodology like I've done here reveals huge errors, almost all of which are designed to give Serral bonus GOAT points. Wouldn't it be better at this point just to admin you've created a Serral = GOAT calculator? It's totally legitimate to call Serral the GOAT--he's probably most people's GOAT, even most players probably think he's the GOAT. You created a highly subjective model to crown Serral the GOAT. It's no different than this GOAT list, just different subjective criteria. It is not rhetorical to say that I penalized Serral, because I did: 1a. Do you deny that the Koreans played more lower tier players in comparison to Serral? Or do you deny that I wrote that I did not include this correction, which is detrimental to Serral? 1b. Do you deny that including non-prime years, which I did, penalizes/hurts Serral the most according to the data? 2. In the article - as well as multiple times in the other thread (and even in the very post you commented on and directly to you in the other thread!!) I explained that this weighting was proposed by ChatGTP. I still don't understand why you keep pounding on that, tbh. As WombaT wrote out in the other thread, the weighting is entirely superfluous. The most important revelation was that most players have mixed results, while Serral is dominating them all. You make it seem like the weighting was there to push Serral, which is not the case. You have no proof for this accusation. On the contrary, I posted the exact text that I used to let ChatGTP decide on the weighting, to let you guys double check. I only did so because I knew based on the results, that the weighting does not change in the first place and I couldn’t decide on a specific weighting, which I explained ad nauseam in the article as well as in the comments. First place would only change with an insane weighting-setup for efficiency… then Life (who most people disregard anyway) would be on top, with Serral being 2nd. Arguing with the weighting result is utterly pointless… As this is your (and other critics) biggest point which you all come back to every time, I doubt that is much more substance to gain from you guys. 3a. In the tournament win %, similar to most metrics, I discounted all region-locked tournaments. This is a penalty to Serral, although a needed one to deny his easier tournament wins. But still it is a penalty; especially to deny them entirely, as Korean pros had rather easy tournament wins as well. 3b. You might disagree that the bonus is not enough according to your subjective feeling. My reasoning for using 20%: I explained in detail why a larger player pool, especially in mostly tier 2, 3 and 4 players, would not decrease a tournament win percentage that much. - doubling the player pool would most likely mean just one extra round. If that round was a group stage, it would naturally mean that you’d have to play more lower tier player statistically, boosting your match win rate - I reasoned that this fact means that average place would need a higher era-multiplier than tournament win percentage - Now looking at Serral’s average win rate of around 82% in his years from 2018-2024 subtracted with 2% (some lower match win for era but also considering that his win rates are deflated versus the Koreans because of 1a, also taking into account that earlier group stages are easier to win than knockout brackets), these more rounds would mean a win probability of roughly 80%. As an example, going from 4 rounds to 5 rounds would statistically mean a win probability 40% to 33%, meaning a loss of 17,5% or a boost of 21,2%. - Now, comparing DH 2020 to a GSL or SSL 2013: Not all tournaments that other players have won were a GSL though. What about the 2013 Hot6ix Cup or ASUS ROG NorthCon? And didn’t Serral win uberly stacked tournaments as well? You are committing a false equivalence/improper comparison fallacy here. In essence, you are right that one can’t compare a GSL to a DH. But that is also not what I did. There were not only GSLs in 2013 (as a matter of fact, there was only one) and Serral won more stacked tournaments as well. It also borders on a strawman, as my approach was not designed to precisely reflect the odds between only one of the hardest 2013 tournaments vs a specific 2020 event, but rather roughly corrects across periods. - Do you further realize that arguing like this from an era/period-perspective also argues against Rogue’s and Maru’s case? - What adjustment would you find appropriate? Overall it is rather the modern Koreans that benefitted in this metric, as Serral, Clem and mostly Reynor were missing in the GSLs that they won. Of course Serral wouldn’t have the same tournament win percentage had he been a full-time pro at his peak playing in the prime era. But the win percentages of INno, Rain, etc. would drop as well, had Mvp, Serral and Clem played there. I explained this already in the article. “Again you call your 20% (or whatever it was) modifier a "penalty" to Serral” Can you give the exact quote? This would be much more fruitful if you didn’t try to rely on notions like my method was not statistical, keep using arguments that have been addressed or straight out deny that Serral was penalized on several occasions, which he clearly was. Talk about the points ratio in the tournaments or the era-multiplier or your preferred weighting. Or - as you by your own account don’t care for statistical GOAT debates anyway - simply stay out of them. But making unsubstantiated accusations like me making a Serral GOAT-calculator or repeating the same already addressed notions is absolutely ridiculous.
You "penalize" Serral within metrics that heavily favor him.
For comparison, it's like if I would incorporate a metric for number of seperate years with a playoffs finish (a metric obviously favoring Maru since he had the longest career), giving it 25% of the overall score, and then acting like I penalized Maru because I considered korean tournaments the same as foreigner-only tournaments, despite being obviously harder.
|
Premo should take their own advice:
On August 10 2024 00:10 PremoBeats wrote:
Seriously.. if there is someone out there taking this obvious bait then it is on them...
|
On July 15 2025 20:14 MJG wrote:Premo should take their own advice: Show nested quote +On August 10 2024 00:10 PremoBeats wrote:
Seriously.. if there is someone out there taking this obvious bait then it is on them...
I don't think Charoisaur and rwala are baiting me from the way they reply... they truly think I favor Serral imo. But perhaps that just went past me, lol.
On July 15 2025 19:57 Charoisaur wrote:Show nested quote +On July 15 2025 19:31 PremoBeats wrote:On July 15 2025 17:32 Charoisaur wrote:On July 15 2025 03:59 rwala wrote:On July 13 2025 21:29 PremoBeats wrote:On July 13 2025 05:44 Charoisaur wrote:On July 12 2025 06:50 Balnazza wrote: but at the same time, it doesn't wank itself into a corner by bias after bias This is literally what you did though? Just to take the WCS eu/am again, it completly fulfills your criteria and yet you disregard it. Yes, the qualifier was open to ALL koreans, they just had to move to Europe or America. That is essentially the same requirement GSL had, you had to live in Korea for a while. The biggest reason not more koreans did it was most likely Proleague, because you couldn't really play Proleague when you played WCS. But that isn't an excluding factor. This was pre-regionlock, when koreans were actually forced to be residents of Europe or America. But the whole criticism before that was how lackluster some koreans were in "their" new region, how little they contributed to the regional scenes. If you want to take prizemoney, just take prizemoney. And yes, then you will have to live with the fact that Clem (by winning EWC), Reynor (by winning Gamers8) and even Maru to a degree (by winning WESG) have a bit of inflated stats, but that is the hand you have been dealt with if you pick prizemoney as the defining statistic. But the second you introduced your "fine-tuning", you made yourself vulnerable...and as you notice, a lot of people (if not all of them) don't agree with your fine-tuning, not even the ones that agree with the outcome. Because it is such a biased and random bunch of multipliers. Just as an example I will point out again the HSC I mentioned earlier, that had barely anyone of notice in attendance but gave out more points than Katowice the next year, just because both tournaments were in different "eras"...divided by two months. damn did the scene drop in two months! And don't get me started on the highly subjective "imbalance"-multiplier... No GOAT list will ever be perfect, that's impossible. If we had like 40 years of SC2 that would have been somewhat the same we would have a good metric to decide by, but considering how much shift and change there has been in the ~15 years lifespan of this game, it will always come down to what statistic you highlight and what metric you choose. But sadly, your way just wasn't solid at all. It was highly biased and too radical to correctly adjust for change (not that you particularly need to). But the funniest part is honestly that you think your approach is completly unbiased... Yeah, this measurement is quite faulty I think, but it brings a great point across which ironically is the one OP quoted in his post "If you torture the data long enough, it will confess to anything" At the end of the day you can use all the fancy stats and data you want, but it doesn't make your ranking any less subjective than someone's guts feeling, whether it's premobeats way overvaluing things like winrates or aligulac rating to prob Serral or the thing ejozl did here. And we all know the one true Goat ranking is this one: 1. Serral 2. Maru (extremely close to Serral) 3. Inno 4. Rogue 5. Life 6. Zest 7. sOs 8. herO 9. Dark 10. Stats I think that quote was put there unironically to highlight the sarcastic over-the-top nature of the article. But why do you mischaracterize what was happening? I wrote in the article that I did not come up with this weighting. I even posted the exact quote that I inserted into ChatGTP so that everyone can check it. Since then, I admitted several times that the weighting is off and even proposed a different one, where Aligulac and match win rates only make up 5% each. Also I never propped Serral lol. He gets penalized the most by subjective decisions I had to make (including non-prime years in tournament win percentage and leave out the inflation-bonus in the match win rate). I remember being asked in the comments of my first article, if I would post the results, if changing the methodology of the original article would mean that Serral would not come out on top in any of the metrics. Well, as it is obvious now, I have no issue accepting that Serral is not the best in tournament win percentage. But it seems like people are trying to critique my result without actually addressing any core methodological feature (except for weighting, which was explained weeks ago). What I would find more troubling than obvious weighting issues is a GOAT ranking that presents itself as statistical but in the end is a post-hoc feeling based scripture. And of course you can make a ranking that is a lot more objective than someone's gut feeling. Not being in alignment with every subjective feeling for era-difference or tournament distinctions does not mean that going as close as possible in a statistical evaluation is the same as a random StarCraft II player that thinks this or that player is the GOAT. We can very clearly see from the statistical evaluation that Serral is way ahead of Maru and Rogue in the time-frame that they played together. We can see that Life was ahead of Rain. We can also see what everyone subjectively thought for a long time: That Rogue has pretty mixed results. Plus, subjectivity can be put into a bare minimum. 1. We need a point/ranking system for the tournaments (similar to what Miz and I did) 2. We need some kind of era-multiplier for different time brackets in terms of how hard it was to win (This can also be calculated more or less ok-ish as well) OR come to the conclusion that all time periods were the same, which I doubt is realistic or correct 3. Decide which metrics are the most important and then add a weighting (the community here and on reddit already agree that tournament score and tournament win participation are the most important These are all the subjective things one needs and the community will probably come to a close proximity. But, even in the case that we have heavy disagreements here: When I revise my method one more time, I will try to make it in a way that people can tell me their desired ratios and I can post the result based on these suggestions. When all my excels are interlinked, I can easily take community suggestions for these three and we can look at how hard we would need to torture the data to make this or that player come out on top. You do not penalize Serral, though I think we all understand rhetorically why you say you do. You give him bonuses. A lot of them. And you only admit it when people call out the more obvious and ridiculous ways in which you do it like your 20% Aligulac weighting. The basic logical error you make is that you are calling all your criteria modifiers "penalties" to Serral but whether they are penalties or bonuses entirely depends on baselines versus expectations in your modifiers. Take your tournament win % criteria. Again you call your 20% (or whatever it was) modifier a "penalty" to Serral. But the question is whether that is too high or too low or just right given that Serral competed exclusively in tournaments that are significantly easier to win than KILs, which have dozens and sometimes hundreds of additional competitors, harder groups stages, more group stages, etc. If it's just right, there's no penalty or bonus to Serral, if it's too high, it's a penalty to Serral, and if it's too low, it's a bonus to Serral. I simply do not agree that the 20% discount reflects the comparative odds of winning a 2020 Dreamhack Masters versus a 2013 GSL or SSL. It's probably 5X easier to win the former than the latter imho. In this specific example, though, it doesn't even matter because it's actually a bonus to Serral irrespective of whether the 20% modifier is right or wrong as he is the only player on your GOAT list that does not have his tournament win % diluted by participation in the signiifcantly-harder-to-win KILs. This is only one example of how you claim to be penalizing Serral when you're actually giving him bonuses in your model. This kind of logical error permeates your model. At first I was giving you the benefit of the doubt that you did not understand some of these concepts in predictive modeling, but it's pretty clear that you do and are just hoping that others do not understand so you can get away with saying over and over again that you are penalizing Serral. Your entire goal is to say that no matter how much you penalize Serral, he is the GOAT, and because you're so focused on this, it's producing many errors and requiring you to go to great lengths to obscure what it is you are doing. The simple fact is that your model giving Serral 2X the GOAT points of the next highest GOAT and 5X the GOAT points of Rogue is on its face not credible. You try to get around this by hiding behind your methodology, but in just a very cursory review of your methodology like I've done here reveals huge errors, almost all of which are designed to give Serral bonus GOAT points. Wouldn't it be better at this point just to admin you've created a Serral = GOAT calculator? It's totally legitimate to call Serral the GOAT--he's probably most people's GOAT, even most players probably think he's the GOAT. You created a highly subjective model to crown Serral the GOAT. It's no different than this GOAT list, just different subjective criteria. Well said rwala's reply consists mostly of straw men, false comparisons and already addressed arguments... what exactly do you think is well said about it? On July 15 2025 03:59 rwala wrote:On July 13 2025 21:29 PremoBeats wrote:On July 13 2025 05:44 Charoisaur wrote:On July 12 2025 06:50 Balnazza wrote: but at the same time, it doesn't wank itself into a corner by bias after bias This is literally what you did though? Just to take the WCS eu/am again, it completly fulfills your criteria and yet you disregard it. Yes, the qualifier was open to ALL koreans, they just had to move to Europe or America. That is essentially the same requirement GSL had, you had to live in Korea for a while. The biggest reason not more koreans did it was most likely Proleague, because you couldn't really play Proleague when you played WCS. But that isn't an excluding factor. This was pre-regionlock, when koreans were actually forced to be residents of Europe or America. But the whole criticism before that was how lackluster some koreans were in "their" new region, how little they contributed to the regional scenes. If you want to take prizemoney, just take prizemoney. And yes, then you will have to live with the fact that Clem (by winning EWC), Reynor (by winning Gamers8) and even Maru to a degree (by winning WESG) have a bit of inflated stats, but that is the hand you have been dealt with if you pick prizemoney as the defining statistic. But the second you introduced your "fine-tuning", you made yourself vulnerable...and as you notice, a lot of people (if not all of them) don't agree with your fine-tuning, not even the ones that agree with the outcome. Because it is such a biased and random bunch of multipliers. Just as an example I will point out again the HSC I mentioned earlier, that had barely anyone of notice in attendance but gave out more points than Katowice the next year, just because both tournaments were in different "eras"...divided by two months. damn did the scene drop in two months! And don't get me started on the highly subjective "imbalance"-multiplier... No GOAT list will ever be perfect, that's impossible. If we had like 40 years of SC2 that would have been somewhat the same we would have a good metric to decide by, but considering how much shift and change there has been in the ~15 years lifespan of this game, it will always come down to what statistic you highlight and what metric you choose. But sadly, your way just wasn't solid at all. It was highly biased and too radical to correctly adjust for change (not that you particularly need to). But the funniest part is honestly that you think your approach is completly unbiased... Yeah, this measurement is quite faulty I think, but it brings a great point across which ironically is the one OP quoted in his post "If you torture the data long enough, it will confess to anything" At the end of the day you can use all the fancy stats and data you want, but it doesn't make your ranking any less subjective than someone's guts feeling, whether it's premobeats way overvaluing things like winrates or aligulac rating to prob Serral or the thing ejozl did here. And we all know the one true Goat ranking is this one: 1. Serral 2. Maru (extremely close to Serral) 3. Inno 4. Rogue 5. Life 6. Zest 7. sOs 8. herO 9. Dark 10. Stats I think that quote was put there unironically to highlight the sarcastic over-the-top nature of the article. But why do you mischaracterize what was happening? I wrote in the article that I did not come up with this weighting. I even posted the exact quote that I inserted into ChatGTP so that everyone can check it. Since then, I admitted several times that the weighting is off and even proposed a different one, where Aligulac and match win rates only make up 5% each. Also I never propped Serral lol. He gets penalized the most by subjective decisions I had to make (including non-prime years in tournament win percentage and leave out the inflation-bonus in the match win rate). I remember being asked in the comments of my first article, if I would post the results, if changing the methodology of the original article would mean that Serral would not come out on top in any of the metrics. Well, as it is obvious now, I have no issue accepting that Serral is not the best in tournament win percentage. But it seems like people are trying to critique my result without actually addressing any core methodological feature (except for weighting, which was explained weeks ago). What I would find more troubling than obvious weighting issues is a GOAT ranking that presents itself as statistical but in the end is a post-hoc feeling based scripture. And of course you can make a ranking that is a lot more objective than someone's gut feeling. Not being in alignment with every subjective feeling for era-difference or tournament distinctions does not mean that going as close as possible in a statistical evaluation is the same as a random StarCraft II player that thinks this or that player is the GOAT. We can very clearly see from the statistical evaluation that Serral is way ahead of Maru and Rogue in the time-frame that they played together. We can see that Life was ahead of Rain. We can also see what everyone subjectively thought for a long time: That Rogue has pretty mixed results. Plus, subjectivity can be put into a bare minimum. 1. We need a point/ranking system for the tournaments (similar to what Miz and I did) 2. We need some kind of era-multiplier for different time brackets in terms of how hard it was to win (This can also be calculated more or less ok-ish as well) OR come to the conclusion that all time periods were the same, which I doubt is realistic or correct 3. Decide which metrics are the most important and then add a weighting (the community here and on reddit already agree that tournament score and tournament win participation are the most important These are all the subjective things one needs and the community will probably come to a close proximity. But, even in the case that we have heavy disagreements here: When I revise my method one more time, I will try to make it in a way that people can tell me their desired ratios and I can post the result based on these suggestions. When all my excels are interlinked, I can easily take community suggestions for these three and we can look at how hard we would need to torture the data to make this or that player come out on top. You do not penalize Serral, though I think we all understand rhetorically why you say you do. You give him bonuses. A lot of them. And you only admit it when people call out the more obvious and ridiculous ways in which you do it like your 20% Aligulac weighting. The basic logical error you make is that you are calling all your criteria modifiers "penalties" to Serral but whether they are penalties or bonuses entirely depends on baselines versus expectations in your modifiers. Take your tournament win % criteria. Again you call your 20% (or whatever it was) modifier a "penalty" to Serral. But the question is whether that is too high or too low or just right given that Serral competed exclusively in tournaments that are significantly easier to win than KILs, which have dozens and sometimes hundreds of additional competitors, harder groups stages, more group stages, etc. If it's just right, there's no penalty or bonus to Serral, if it's too high, it's a penalty to Serral, and if it's too low, it's a bonus to Serral. I simply do not agree that the 20% discount reflects the comparative odds of winning a 2020 Dreamhack Masters versus a 2013 GSL or SSL. It's probably 5X easier to win the former than the latter imho. In this specific example, though, it doesn't even matter because it's actually a bonus to Serral irrespective of whether the 20% modifier is right or wrong as he is the only player on your GOAT list that does not have his tournament win % diluted by participation in the signiifcantly-harder-to-win KILs. This is only one example of how you claim to be penalizing Serral when you're actually giving him bonuses in your model. This kind of logical error permeates your model. At first I was giving you the benefit of the doubt that you did not understand some of these concepts in predictive modeling, but it's pretty clear that you do and are just hoping that others do not understand so you can get away with saying over and over again that you are penalizing Serral. Your entire goal is to say that no matter how much you penalize Serral, he is the GOAT, and because you're so focused on this, it's producing many errors and requiring you to go to great lengths to obscure what it is you are doing. The simple fact is that your model giving Serral 2X the GOAT points of the next highest GOAT and 5X the GOAT points of Rogue is on its face not credible. You try to get around this by hiding behind your methodology, but in just a very cursory review of your methodology like I've done here reveals huge errors, almost all of which are designed to give Serral bonus GOAT points. Wouldn't it be better at this point just to admin you've created a Serral = GOAT calculator? It's totally legitimate to call Serral the GOAT--he's probably most people's GOAT, even most players probably think he's the GOAT. You created a highly subjective model to crown Serral the GOAT. It's no different than this GOAT list, just different subjective criteria. It is not rhetorical to say that I penalized Serral, because I did: 1a. Do you deny that the Koreans played more lower tier players in comparison to Serral? Or do you deny that I wrote that I did not include this correction, which is detrimental to Serral? 1b. Do you deny that including non-prime years, which I did, penalizes/hurts Serral the most according to the data? 2. In the article - as well as multiple times in the other thread (and even in the very post you commented on and directly to you in the other thread!!) I explained that this weighting was proposed by ChatGTP. I still don't understand why you keep pounding on that, tbh. As WombaT wrote out in the other thread, the weighting is entirely superfluous. The most important revelation was that most players have mixed results, while Serral is dominating them all. You make it seem like the weighting was there to push Serral, which is not the case. You have no proof for this accusation. On the contrary, I posted the exact text that I used to let ChatGTP decide on the weighting, to let you guys double check. I only did so because I knew based on the results, that the weighting does not change in the first place and I couldn’t decide on a specific weighting, which I explained ad nauseam in the article as well as in the comments. First place would only change with an insane weighting-setup for efficiency… then Life (who most people disregard anyway) would be on top, with Serral being 2nd. Arguing with the weighting result is utterly pointless… As this is your (and other critics) biggest point which you all come back to every time, I doubt that is much more substance to gain from you guys. 3a. In the tournament win %, similar to most metrics, I discounted all region-locked tournaments. This is a penalty to Serral, although a needed one to deny his easier tournament wins. But still it is a penalty; especially to deny them entirely, as Korean pros had rather easy tournament wins as well. 3b. You might disagree that the bonus is not enough according to your subjective feeling. My reasoning for using 20%: I explained in detail why a larger player pool, especially in mostly tier 2, 3 and 4 players, would not decrease a tournament win percentage that much. - doubling the player pool would most likely mean just one extra round. If that round was a group stage, it would naturally mean that you’d have to play more lower tier player statistically, boosting your match win rate - I reasoned that this fact means that average place would need a higher era-multiplier than tournament win percentage - Now looking at Serral’s average win rate of around 82% in his years from 2018-2024 subtracted with 2% (some lower match win for era but also considering that his win rates are deflated versus the Koreans because of 1a, also taking into account that earlier group stages are easier to win than knockout brackets), these more rounds would mean a win probability of roughly 80%. As an example, going from 4 rounds to 5 rounds would statistically mean a win probability 40% to 33%, meaning a loss of 17,5% or a boost of 21,2%. - Now, comparing DH 2020 to a GSL or SSL 2013: Not all tournaments that other players have won were a GSL though. What about the 2013 Hot6ix Cup or ASUS ROG NorthCon? And didn’t Serral win uberly stacked tournaments as well? You are committing a false equivalence/improper comparison fallacy here. In essence, you are right that one can’t compare a GSL to a DH. But that is also not what I did. There were not only GSLs in 2013 (as a matter of fact, there was only one) and Serral won more stacked tournaments as well. It also borders on a strawman, as my approach was not designed to precisely reflect the odds between only one of the hardest 2013 tournaments vs a specific 2020 event, but rather roughly corrects across periods. - Do you further realize that arguing like this from an era/period-perspective also argues against Rogue’s and Maru’s case? - What adjustment would you find appropriate? Overall it is rather the modern Koreans that benefitted in this metric, as Serral, Clem and mostly Reynor were missing in the GSLs that they won. Of course Serral wouldn’t have the same tournament win percentage had he been a full-time pro at his peak playing in the prime era. But the win percentages of INno, Rain, etc. would drop as well, had Mvp, Serral and Clem played there. I explained this already in the article. “Again you call your 20% (or whatever it was) modifier a "penalty" to Serral” Can you give the exact quote? This would be much more fruitful if you didn’t try to rely on notions like my method was not statistical, keep using arguments that have been addressed or straight out deny that Serral was penalized on several occasions, which he clearly was. Talk about the points ratio in the tournaments or the era-multiplier or your preferred weighting. Or - as you by your own account don’t care for statistical GOAT debates anyway - simply stay out of them. But making unsubstantiated accusations like me making a Serral GOAT-calculator or repeating the same already addressed notions is absolutely ridiculous. You "penalize" Serral within metrics that heavily favor him. For comparison, it's like if I would incorporate a metric for number of seperate years with a playoffs finish (a metric obviously favoring Maru since he had the longest career), giving it 25% of the overall score, and then acting like I penalized Maru because I considered korean tournaments the same as foreigner-only tournaments, despite being obviously harder.
I already said in the article that there are a million ways to fragment the data... one could also make a scoreboard for best cannon rush defender or what you just suggested. But it simply makes no sense, as I used data that covers the entirety of the game and metrics that are most valuable to determine the Greatest of All Time. If you disagree, simply explain why your example is a worthy metric.
But your comparison is completely off, as the tournament win percentage is a simple data set. Nothing of it has been modified like you did. The only thing that was modified was
- disregarding region locks, which is a (needed) penalty for Serral
- add a multiplier that boosted everything pre-2018 and penalized everything post-2018 (which penalized Serral among others)
So which metric does favor Serral? And why exactly does it "favor" him?
|
On July 15 2025 20:20 PremoBeats wrote:Show nested quote +On July 15 2025 20:14 MJG wrote:Premo should take their own advice: On August 10 2024 00:10 PremoBeats wrote:
Seriously.. if there is someone out there taking this obvious bait then it is on them... I don't think Charoisaur and rwala are baiting me from the way they reply... they truly think I favor Serral imo. But perhaps that just went past me, lol. Show nested quote +On July 15 2025 19:57 Charoisaur wrote:On July 15 2025 19:31 PremoBeats wrote:On July 15 2025 17:32 Charoisaur wrote:On July 15 2025 03:59 rwala wrote:On July 13 2025 21:29 PremoBeats wrote:On July 13 2025 05:44 Charoisaur wrote:On July 12 2025 06:50 Balnazza wrote: but at the same time, it doesn't wank itself into a corner by bias after bias This is literally what you did though? Just to take the WCS eu/am again, it completly fulfills your criteria and yet you disregard it. Yes, the qualifier was open to ALL koreans, they just had to move to Europe or America. That is essentially the same requirement GSL had, you had to live in Korea for a while. The biggest reason not more koreans did it was most likely Proleague, because you couldn't really play Proleague when you played WCS. But that isn't an excluding factor. This was pre-regionlock, when koreans were actually forced to be residents of Europe or America. But the whole criticism before that was how lackluster some koreans were in "their" new region, how little they contributed to the regional scenes. If you want to take prizemoney, just take prizemoney. And yes, then you will have to live with the fact that Clem (by winning EWC), Reynor (by winning Gamers8) and even Maru to a degree (by winning WESG) have a bit of inflated stats, but that is the hand you have been dealt with if you pick prizemoney as the defining statistic. But the second you introduced your "fine-tuning", you made yourself vulnerable...and as you notice, a lot of people (if not all of them) don't agree with your fine-tuning, not even the ones that agree with the outcome. Because it is such a biased and random bunch of multipliers. Just as an example I will point out again the HSC I mentioned earlier, that had barely anyone of notice in attendance but gave out more points than Katowice the next year, just because both tournaments were in different "eras"...divided by two months. damn did the scene drop in two months! And don't get me started on the highly subjective "imbalance"-multiplier... No GOAT list will ever be perfect, that's impossible. If we had like 40 years of SC2 that would have been somewhat the same we would have a good metric to decide by, but considering how much shift and change there has been in the ~15 years lifespan of this game, it will always come down to what statistic you highlight and what metric you choose. But sadly, your way just wasn't solid at all. It was highly biased and too radical to correctly adjust for change (not that you particularly need to). But the funniest part is honestly that you think your approach is completly unbiased... Yeah, this measurement is quite faulty I think, but it brings a great point across which ironically is the one OP quoted in his post "If you torture the data long enough, it will confess to anything" At the end of the day you can use all the fancy stats and data you want, but it doesn't make your ranking any less subjective than someone's guts feeling, whether it's premobeats way overvaluing things like winrates or aligulac rating to prob Serral or the thing ejozl did here. And we all know the one true Goat ranking is this one: 1. Serral 2. Maru (extremely close to Serral) 3. Inno 4. Rogue 5. Life 6. Zest 7. sOs 8. herO 9. Dark 10. Stats I think that quote was put there unironically to highlight the sarcastic over-the-top nature of the article. But why do you mischaracterize what was happening? I wrote in the article that I did not come up with this weighting. I even posted the exact quote that I inserted into ChatGTP so that everyone can check it. Since then, I admitted several times that the weighting is off and even proposed a different one, where Aligulac and match win rates only make up 5% each. Also I never propped Serral lol. He gets penalized the most by subjective decisions I had to make (including non-prime years in tournament win percentage and leave out the inflation-bonus in the match win rate). I remember being asked in the comments of my first article, if I would post the results, if changing the methodology of the original article would mean that Serral would not come out on top in any of the metrics. Well, as it is obvious now, I have no issue accepting that Serral is not the best in tournament win percentage. But it seems like people are trying to critique my result without actually addressing any core methodological feature (except for weighting, which was explained weeks ago). What I would find more troubling than obvious weighting issues is a GOAT ranking that presents itself as statistical but in the end is a post-hoc feeling based scripture. And of course you can make a ranking that is a lot more objective than someone's gut feeling. Not being in alignment with every subjective feeling for era-difference or tournament distinctions does not mean that going as close as possible in a statistical evaluation is the same as a random StarCraft II player that thinks this or that player is the GOAT. We can very clearly see from the statistical evaluation that Serral is way ahead of Maru and Rogue in the time-frame that they played together. We can see that Life was ahead of Rain. We can also see what everyone subjectively thought for a long time: That Rogue has pretty mixed results. Plus, subjectivity can be put into a bare minimum. 1. We need a point/ranking system for the tournaments (similar to what Miz and I did) 2. We need some kind of era-multiplier for different time brackets in terms of how hard it was to win (This can also be calculated more or less ok-ish as well) OR come to the conclusion that all time periods were the same, which I doubt is realistic or correct 3. Decide which metrics are the most important and then add a weighting (the community here and on reddit already agree that tournament score and tournament win participation are the most important These are all the subjective things one needs and the community will probably come to a close proximity. But, even in the case that we have heavy disagreements here: When I revise my method one more time, I will try to make it in a way that people can tell me their desired ratios and I can post the result based on these suggestions. When all my excels are interlinked, I can easily take community suggestions for these three and we can look at how hard we would need to torture the data to make this or that player come out on top. You do not penalize Serral, though I think we all understand rhetorically why you say you do. You give him bonuses. A lot of them. And you only admit it when people call out the more obvious and ridiculous ways in which you do it like your 20% Aligulac weighting. The basic logical error you make is that you are calling all your criteria modifiers "penalties" to Serral but whether they are penalties or bonuses entirely depends on baselines versus expectations in your modifiers. Take your tournament win % criteria. Again you call your 20% (or whatever it was) modifier a "penalty" to Serral. But the question is whether that is too high or too low or just right given that Serral competed exclusively in tournaments that are significantly easier to win than KILs, which have dozens and sometimes hundreds of additional competitors, harder groups stages, more group stages, etc. If it's just right, there's no penalty or bonus to Serral, if it's too high, it's a penalty to Serral, and if it's too low, it's a bonus to Serral. I simply do not agree that the 20% discount reflects the comparative odds of winning a 2020 Dreamhack Masters versus a 2013 GSL or SSL. It's probably 5X easier to win the former than the latter imho. In this specific example, though, it doesn't even matter because it's actually a bonus to Serral irrespective of whether the 20% modifier is right or wrong as he is the only player on your GOAT list that does not have his tournament win % diluted by participation in the signiifcantly-harder-to-win KILs. This is only one example of how you claim to be penalizing Serral when you're actually giving him bonuses in your model. This kind of logical error permeates your model. At first I was giving you the benefit of the doubt that you did not understand some of these concepts in predictive modeling, but it's pretty clear that you do and are just hoping that others do not understand so you can get away with saying over and over again that you are penalizing Serral. Your entire goal is to say that no matter how much you penalize Serral, he is the GOAT, and because you're so focused on this, it's producing many errors and requiring you to go to great lengths to obscure what it is you are doing. The simple fact is that your model giving Serral 2X the GOAT points of the next highest GOAT and 5X the GOAT points of Rogue is on its face not credible. You try to get around this by hiding behind your methodology, but in just a very cursory review of your methodology like I've done here reveals huge errors, almost all of which are designed to give Serral bonus GOAT points. Wouldn't it be better at this point just to admin you've created a Serral = GOAT calculator? It's totally legitimate to call Serral the GOAT--he's probably most people's GOAT, even most players probably think he's the GOAT. You created a highly subjective model to crown Serral the GOAT. It's no different than this GOAT list, just different subjective criteria. Well said rwala's reply consists mostly of straw men, false comparisons and already addressed arguments... what exactly do you think is well said about it? On July 15 2025 03:59 rwala wrote:On July 13 2025 21:29 PremoBeats wrote:On July 13 2025 05:44 Charoisaur wrote:On July 12 2025 06:50 Balnazza wrote: but at the same time, it doesn't wank itself into a corner by bias after bias This is literally what you did though? Just to take the WCS eu/am again, it completly fulfills your criteria and yet you disregard it. Yes, the qualifier was open to ALL koreans, they just had to move to Europe or America. That is essentially the same requirement GSL had, you had to live in Korea for a while. The biggest reason not more koreans did it was most likely Proleague, because you couldn't really play Proleague when you played WCS. But that isn't an excluding factor. This was pre-regionlock, when koreans were actually forced to be residents of Europe or America. But the whole criticism before that was how lackluster some koreans were in "their" new region, how little they contributed to the regional scenes. If you want to take prizemoney, just take prizemoney. And yes, then you will have to live with the fact that Clem (by winning EWC), Reynor (by winning Gamers8) and even Maru to a degree (by winning WESG) have a bit of inflated stats, but that is the hand you have been dealt with if you pick prizemoney as the defining statistic. But the second you introduced your "fine-tuning", you made yourself vulnerable...and as you notice, a lot of people (if not all of them) don't agree with your fine-tuning, not even the ones that agree with the outcome. Because it is such a biased and random bunch of multipliers. Just as an example I will point out again the HSC I mentioned earlier, that had barely anyone of notice in attendance but gave out more points than Katowice the next year, just because both tournaments were in different "eras"...divided by two months. damn did the scene drop in two months! And don't get me started on the highly subjective "imbalance"-multiplier... No GOAT list will ever be perfect, that's impossible. If we had like 40 years of SC2 that would have been somewhat the same we would have a good metric to decide by, but considering how much shift and change there has been in the ~15 years lifespan of this game, it will always come down to what statistic you highlight and what metric you choose. But sadly, your way just wasn't solid at all. It was highly biased and too radical to correctly adjust for change (not that you particularly need to). But the funniest part is honestly that you think your approach is completly unbiased... Yeah, this measurement is quite faulty I think, but it brings a great point across which ironically is the one OP quoted in his post "If you torture the data long enough, it will confess to anything" At the end of the day you can use all the fancy stats and data you want, but it doesn't make your ranking any less subjective than someone's guts feeling, whether it's premobeats way overvaluing things like winrates or aligulac rating to prob Serral or the thing ejozl did here. And we all know the one true Goat ranking is this one: 1. Serral 2. Maru (extremely close to Serral) 3. Inno 4. Rogue 5. Life 6. Zest 7. sOs 8. herO 9. Dark 10. Stats I think that quote was put there unironically to highlight the sarcastic over-the-top nature of the article. But why do you mischaracterize what was happening? I wrote in the article that I did not come up with this weighting. I even posted the exact quote that I inserted into ChatGTP so that everyone can check it. Since then, I admitted several times that the weighting is off and even proposed a different one, where Aligulac and match win rates only make up 5% each. Also I never propped Serral lol. He gets penalized the most by subjective decisions I had to make (including non-prime years in tournament win percentage and leave out the inflation-bonus in the match win rate). I remember being asked in the comments of my first article, if I would post the results, if changing the methodology of the original article would mean that Serral would not come out on top in any of the metrics. Well, as it is obvious now, I have no issue accepting that Serral is not the best in tournament win percentage. But it seems like people are trying to critique my result without actually addressing any core methodological feature (except for weighting, which was explained weeks ago). What I would find more troubling than obvious weighting issues is a GOAT ranking that presents itself as statistical but in the end is a post-hoc feeling based scripture. And of course you can make a ranking that is a lot more objective than someone's gut feeling. Not being in alignment with every subjective feeling for era-difference or tournament distinctions does not mean that going as close as possible in a statistical evaluation is the same as a random StarCraft II player that thinks this or that player is the GOAT. We can very clearly see from the statistical evaluation that Serral is way ahead of Maru and Rogue in the time-frame that they played together. We can see that Life was ahead of Rain. We can also see what everyone subjectively thought for a long time: That Rogue has pretty mixed results. Plus, subjectivity can be put into a bare minimum. 1. We need a point/ranking system for the tournaments (similar to what Miz and I did) 2. We need some kind of era-multiplier for different time brackets in terms of how hard it was to win (This can also be calculated more or less ok-ish as well) OR come to the conclusion that all time periods were the same, which I doubt is realistic or correct 3. Decide which metrics are the most important and then add a weighting (the community here and on reddit already agree that tournament score and tournament win participation are the most important These are all the subjective things one needs and the community will probably come to a close proximity. But, even in the case that we have heavy disagreements here: When I revise my method one more time, I will try to make it in a way that people can tell me their desired ratios and I can post the result based on these suggestions. When all my excels are interlinked, I can easily take community suggestions for these three and we can look at how hard we would need to torture the data to make this or that player come out on top. You do not penalize Serral, though I think we all understand rhetorically why you say you do. You give him bonuses. A lot of them. And you only admit it when people call out the more obvious and ridiculous ways in which you do it like your 20% Aligulac weighting. The basic logical error you make is that you are calling all your criteria modifiers "penalties" to Serral but whether they are penalties or bonuses entirely depends on baselines versus expectations in your modifiers. Take your tournament win % criteria. Again you call your 20% (or whatever it was) modifier a "penalty" to Serral. But the question is whether that is too high or too low or just right given that Serral competed exclusively in tournaments that are significantly easier to win than KILs, which have dozens and sometimes hundreds of additional competitors, harder groups stages, more group stages, etc. If it's just right, there's no penalty or bonus to Serral, if it's too high, it's a penalty to Serral, and if it's too low, it's a bonus to Serral. I simply do not agree that the 20% discount reflects the comparative odds of winning a 2020 Dreamhack Masters versus a 2013 GSL or SSL. It's probably 5X easier to win the former than the latter imho. In this specific example, though, it doesn't even matter because it's actually a bonus to Serral irrespective of whether the 20% modifier is right or wrong as he is the only player on your GOAT list that does not have his tournament win % diluted by participation in the signiifcantly-harder-to-win KILs. This is only one example of how you claim to be penalizing Serral when you're actually giving him bonuses in your model. This kind of logical error permeates your model. At first I was giving you the benefit of the doubt that you did not understand some of these concepts in predictive modeling, but it's pretty clear that you do and are just hoping that others do not understand so you can get away with saying over and over again that you are penalizing Serral. Your entire goal is to say that no matter how much you penalize Serral, he is the GOAT, and because you're so focused on this, it's producing many errors and requiring you to go to great lengths to obscure what it is you are doing. The simple fact is that your model giving Serral 2X the GOAT points of the next highest GOAT and 5X the GOAT points of Rogue is on its face not credible. You try to get around this by hiding behind your methodology, but in just a very cursory review of your methodology like I've done here reveals huge errors, almost all of which are designed to give Serral bonus GOAT points. Wouldn't it be better at this point just to admin you've created a Serral = GOAT calculator? It's totally legitimate to call Serral the GOAT--he's probably most people's GOAT, even most players probably think he's the GOAT. You created a highly subjective model to crown Serral the GOAT. It's no different than this GOAT list, just different subjective criteria. It is not rhetorical to say that I penalized Serral, because I did: 1a. Do you deny that the Koreans played more lower tier players in comparison to Serral? Or do you deny that I wrote that I did not include this correction, which is detrimental to Serral? 1b. Do you deny that including non-prime years, which I did, penalizes/hurts Serral the most according to the data? 2. In the article - as well as multiple times in the other thread (and even in the very post you commented on and directly to you in the other thread!!) I explained that this weighting was proposed by ChatGTP. I still don't understand why you keep pounding on that, tbh. As WombaT wrote out in the other thread, the weighting is entirely superfluous. The most important revelation was that most players have mixed results, while Serral is dominating them all. You make it seem like the weighting was there to push Serral, which is not the case. You have no proof for this accusation. On the contrary, I posted the exact text that I used to let ChatGTP decide on the weighting, to let you guys double check. I only did so because I knew based on the results, that the weighting does not change in the first place and I couldn’t decide on a specific weighting, which I explained ad nauseam in the article as well as in the comments. First place would only change with an insane weighting-setup for efficiency… then Life (who most people disregard anyway) would be on top, with Serral being 2nd. Arguing with the weighting result is utterly pointless… As this is your (and other critics) biggest point which you all come back to every time, I doubt that is much more substance to gain from you guys. 3a. In the tournament win %, similar to most metrics, I discounted all region-locked tournaments. This is a penalty to Serral, although a needed one to deny his easier tournament wins. But still it is a penalty; especially to deny them entirely, as Korean pros had rather easy tournament wins as well. 3b. You might disagree that the bonus is not enough according to your subjective feeling. My reasoning for using 20%: I explained in detail why a larger player pool, especially in mostly tier 2, 3 and 4 players, would not decrease a tournament win percentage that much. - doubling the player pool would most likely mean just one extra round. If that round was a group stage, it would naturally mean that you’d have to play more lower tier player statistically, boosting your match win rate - I reasoned that this fact means that average place would need a higher era-multiplier than tournament win percentage - Now looking at Serral’s average win rate of around 82% in his years from 2018-2024 subtracted with 2% (some lower match win for era but also considering that his win rates are deflated versus the Koreans because of 1a, also taking into account that earlier group stages are easier to win than knockout brackets), these more rounds would mean a win probability of roughly 80%. As an example, going from 4 rounds to 5 rounds would statistically mean a win probability 40% to 33%, meaning a loss of 17,5% or a boost of 21,2%. - Now, comparing DH 2020 to a GSL or SSL 2013: Not all tournaments that other players have won were a GSL though. What about the 2013 Hot6ix Cup or ASUS ROG NorthCon? And didn’t Serral win uberly stacked tournaments as well? You are committing a false equivalence/improper comparison fallacy here. In essence, you are right that one can’t compare a GSL to a DH. But that is also not what I did. There were not only GSLs in 2013 (as a matter of fact, there was only one) and Serral won more stacked tournaments as well. It also borders on a strawman, as my approach was not designed to precisely reflect the odds between only one of the hardest 2013 tournaments vs a specific 2020 event, but rather roughly corrects across periods. - Do you further realize that arguing like this from an era/period-perspective also argues against Rogue’s and Maru’s case? - What adjustment would you find appropriate? Overall it is rather the modern Koreans that benefitted in this metric, as Serral, Clem and mostly Reynor were missing in the GSLs that they won. Of course Serral wouldn’t have the same tournament win percentage had he been a full-time pro at his peak playing in the prime era. But the win percentages of INno, Rain, etc. would drop as well, had Mvp, Serral and Clem played there. I explained this already in the article. “Again you call your 20% (or whatever it was) modifier a "penalty" to Serral” Can you give the exact quote? This would be much more fruitful if you didn’t try to rely on notions like my method was not statistical, keep using arguments that have been addressed or straight out deny that Serral was penalized on several occasions, which he clearly was. Talk about the points ratio in the tournaments or the era-multiplier or your preferred weighting. Or - as you by your own account don’t care for statistical GOAT debates anyway - simply stay out of them. But making unsubstantiated accusations like me making a Serral GOAT-calculator or repeating the same already addressed notions is absolutely ridiculous. You "penalize" Serral within metrics that heavily favor him. For comparison, it's like if I would incorporate a metric for number of seperate years with a playoffs finish (a metric obviously favoring Maru since he had the longest career), giving it 25% of the overall score, and then acting like I penalized Maru because I considered korean tournaments the same as foreigner-only tournaments, despite being obviously harder. Which metric does favor Serral? And why does it "favor" him? Because he is good in it?
Aligulac rating, match winrate, tournament winrate, average placing are all metrics that favor Serral. Yes, because he's good at it, but you attributed all those metrics an unreasonably high influence on the final score, especially because some of them measure very similar things and thus have massive collinearity e.g. aligulac rating and average placement.
Metrics in which Serral wouldn't perform as well however are excluded, like longevity focused metrics, e.g. the aforementioned number of seperate years with a top x finish.
|
On July 15 2025 20:20 PremoBeats wrote:Show nested quote +On July 15 2025 20:14 MJG wrote:Premo should take their own advice: On August 10 2024 00:10 PremoBeats wrote:
Seriously.. if there is someone out there taking this obvious bait then it is on them... I don't think Charoisaur and rwala are baiting me from the way they reply... they truly think I favor Serral imo. But perhaps that just went past me, lol. Show nested quote +On July 15 2025 19:57 Charoisaur wrote:On July 15 2025 19:31 PremoBeats wrote:On July 15 2025 17:32 Charoisaur wrote:On July 15 2025 03:59 rwala wrote:On July 13 2025 21:29 PremoBeats wrote:On July 13 2025 05:44 Charoisaur wrote:On July 12 2025 06:50 Balnazza wrote: but at the same time, it doesn't wank itself into a corner by bias after bias This is literally what you did though? Just to take the WCS eu/am again, it completly fulfills your criteria and yet you disregard it. Yes, the qualifier was open to ALL koreans, they just had to move to Europe or America. That is essentially the same requirement GSL had, you had to live in Korea for a while. The biggest reason not more koreans did it was most likely Proleague, because you couldn't really play Proleague when you played WCS. But that isn't an excluding factor. This was pre-regionlock, when koreans were actually forced to be residents of Europe or America. But the whole criticism before that was how lackluster some koreans were in "their" new region, how little they contributed to the regional scenes. If you want to take prizemoney, just take prizemoney. And yes, then you will have to live with the fact that Clem (by winning EWC), Reynor (by winning Gamers8) and even Maru to a degree (by winning WESG) have a bit of inflated stats, but that is the hand you have been dealt with if you pick prizemoney as the defining statistic. But the second you introduced your "fine-tuning", you made yourself vulnerable...and as you notice, a lot of people (if not all of them) don't agree with your fine-tuning, not even the ones that agree with the outcome. Because it is such a biased and random bunch of multipliers. Just as an example I will point out again the HSC I mentioned earlier, that had barely anyone of notice in attendance but gave out more points than Katowice the next year, just because both tournaments were in different "eras"...divided by two months. damn did the scene drop in two months! And don't get me started on the highly subjective "imbalance"-multiplier... No GOAT list will ever be perfect, that's impossible. If we had like 40 years of SC2 that would have been somewhat the same we would have a good metric to decide by, but considering how much shift and change there has been in the ~15 years lifespan of this game, it will always come down to what statistic you highlight and what metric you choose. But sadly, your way just wasn't solid at all. It was highly biased and too radical to correctly adjust for change (not that you particularly need to). But the funniest part is honestly that you think your approach is completly unbiased... Yeah, this measurement is quite faulty I think, but it brings a great point across which ironically is the one OP quoted in his post "If you torture the data long enough, it will confess to anything" At the end of the day you can use all the fancy stats and data you want, but it doesn't make your ranking any less subjective than someone's guts feeling, whether it's premobeats way overvaluing things like winrates or aligulac rating to prob Serral or the thing ejozl did here. And we all know the one true Goat ranking is this one: 1. Serral 2. Maru (extremely close to Serral) 3. Inno 4. Rogue 5. Life 6. Zest 7. sOs 8. herO 9. Dark 10. Stats I think that quote was put there unironically to highlight the sarcastic over-the-top nature of the article. But why do you mischaracterize what was happening? I wrote in the article that I did not come up with this weighting. I even posted the exact quote that I inserted into ChatGTP so that everyone can check it. Since then, I admitted several times that the weighting is off and even proposed a different one, where Aligulac and match win rates only make up 5% each. Also I never propped Serral lol. He gets penalized the most by subjective decisions I had to make (including non-prime years in tournament win percentage and leave out the inflation-bonus in the match win rate). I remember being asked in the comments of my first article, if I would post the results, if changing the methodology of the original article would mean that Serral would not come out on top in any of the metrics. Well, as it is obvious now, I have no issue accepting that Serral is not the best in tournament win percentage. But it seems like people are trying to critique my result without actually addressing any core methodological feature (except for weighting, which was explained weeks ago). What I would find more troubling than obvious weighting issues is a GOAT ranking that presents itself as statistical but in the end is a post-hoc feeling based scripture. And of course you can make a ranking that is a lot more objective than someone's gut feeling. Not being in alignment with every subjective feeling for era-difference or tournament distinctions does not mean that going as close as possible in a statistical evaluation is the same as a random StarCraft II player that thinks this or that player is the GOAT. We can very clearly see from the statistical evaluation that Serral is way ahead of Maru and Rogue in the time-frame that they played together. We can see that Life was ahead of Rain. We can also see what everyone subjectively thought for a long time: That Rogue has pretty mixed results. Plus, subjectivity can be put into a bare minimum. 1. We need a point/ranking system for the tournaments (similar to what Miz and I did) 2. We need some kind of era-multiplier for different time brackets in terms of how hard it was to win (This can also be calculated more or less ok-ish as well) OR come to the conclusion that all time periods were the same, which I doubt is realistic or correct 3. Decide which metrics are the most important and then add a weighting (the community here and on reddit already agree that tournament score and tournament win participation are the most important These are all the subjective things one needs and the community will probably come to a close proximity. But, even in the case that we have heavy disagreements here: When I revise my method one more time, I will try to make it in a way that people can tell me their desired ratios and I can post the result based on these suggestions. When all my excels are interlinked, I can easily take community suggestions for these three and we can look at how hard we would need to torture the data to make this or that player come out on top. You do not penalize Serral, though I think we all understand rhetorically why you say you do. You give him bonuses. A lot of them. And you only admit it when people call out the more obvious and ridiculous ways in which you do it like your 20% Aligulac weighting. The basic logical error you make is that you are calling all your criteria modifiers "penalties" to Serral but whether they are penalties or bonuses entirely depends on baselines versus expectations in your modifiers. Take your tournament win % criteria. Again you call your 20% (or whatever it was) modifier a "penalty" to Serral. But the question is whether that is too high or too low or just right given that Serral competed exclusively in tournaments that are significantly easier to win than KILs, which have dozens and sometimes hundreds of additional competitors, harder groups stages, more group stages, etc. If it's just right, there's no penalty or bonus to Serral, if it's too high, it's a penalty to Serral, and if it's too low, it's a bonus to Serral. I simply do not agree that the 20% discount reflects the comparative odds of winning a 2020 Dreamhack Masters versus a 2013 GSL or SSL. It's probably 5X easier to win the former than the latter imho. In this specific example, though, it doesn't even matter because it's actually a bonus to Serral irrespective of whether the 20% modifier is right or wrong as he is the only player on your GOAT list that does not have his tournament win % diluted by participation in the signiifcantly-harder-to-win KILs. This is only one example of how you claim to be penalizing Serral when you're actually giving him bonuses in your model. This kind of logical error permeates your model. At first I was giving you the benefit of the doubt that you did not understand some of these concepts in predictive modeling, but it's pretty clear that you do and are just hoping that others do not understand so you can get away with saying over and over again that you are penalizing Serral. Your entire goal is to say that no matter how much you penalize Serral, he is the GOAT, and because you're so focused on this, it's producing many errors and requiring you to go to great lengths to obscure what it is you are doing. The simple fact is that your model giving Serral 2X the GOAT points of the next highest GOAT and 5X the GOAT points of Rogue is on its face not credible. You try to get around this by hiding behind your methodology, but in just a very cursory review of your methodology like I've done here reveals huge errors, almost all of which are designed to give Serral bonus GOAT points. Wouldn't it be better at this point just to admin you've created a Serral = GOAT calculator? It's totally legitimate to call Serral the GOAT--he's probably most people's GOAT, even most players probably think he's the GOAT. You created a highly subjective model to crown Serral the GOAT. It's no different than this GOAT list, just different subjective criteria. Well said rwala's reply consists mostly of straw men, false comparisons and already addressed arguments... what exactly do you think is well said about it? On July 15 2025 03:59 rwala wrote:On July 13 2025 21:29 PremoBeats wrote:On July 13 2025 05:44 Charoisaur wrote:On July 12 2025 06:50 Balnazza wrote: but at the same time, it doesn't wank itself into a corner by bias after bias This is literally what you did though? Just to take the WCS eu/am again, it completly fulfills your criteria and yet you disregard it. Yes, the qualifier was open to ALL koreans, they just had to move to Europe or America. That is essentially the same requirement GSL had, you had to live in Korea for a while. The biggest reason not more koreans did it was most likely Proleague, because you couldn't really play Proleague when you played WCS. But that isn't an excluding factor. This was pre-regionlock, when koreans were actually forced to be residents of Europe or America. But the whole criticism before that was how lackluster some koreans were in "their" new region, how little they contributed to the regional scenes. If you want to take prizemoney, just take prizemoney. And yes, then you will have to live with the fact that Clem (by winning EWC), Reynor (by winning Gamers8) and even Maru to a degree (by winning WESG) have a bit of inflated stats, but that is the hand you have been dealt with if you pick prizemoney as the defining statistic. But the second you introduced your "fine-tuning", you made yourself vulnerable...and as you notice, a lot of people (if not all of them) don't agree with your fine-tuning, not even the ones that agree with the outcome. Because it is such a biased and random bunch of multipliers. Just as an example I will point out again the HSC I mentioned earlier, that had barely anyone of notice in attendance but gave out more points than Katowice the next year, just because both tournaments were in different "eras"...divided by two months. damn did the scene drop in two months! And don't get me started on the highly subjective "imbalance"-multiplier... No GOAT list will ever be perfect, that's impossible. If we had like 40 years of SC2 that would have been somewhat the same we would have a good metric to decide by, but considering how much shift and change there has been in the ~15 years lifespan of this game, it will always come down to what statistic you highlight and what metric you choose. But sadly, your way just wasn't solid at all. It was highly biased and too radical to correctly adjust for change (not that you particularly need to). But the funniest part is honestly that you think your approach is completly unbiased... Yeah, this measurement is quite faulty I think, but it brings a great point across which ironically is the one OP quoted in his post "If you torture the data long enough, it will confess to anything" At the end of the day you can use all the fancy stats and data you want, but it doesn't make your ranking any less subjective than someone's guts feeling, whether it's premobeats way overvaluing things like winrates or aligulac rating to prob Serral or the thing ejozl did here. And we all know the one true Goat ranking is this one: 1. Serral 2. Maru (extremely close to Serral) 3. Inno 4. Rogue 5. Life 6. Zest 7. sOs 8. herO 9. Dark 10. Stats I think that quote was put there unironically to highlight the sarcastic over-the-top nature of the article. But why do you mischaracterize what was happening? I wrote in the article that I did not come up with this weighting. I even posted the exact quote that I inserted into ChatGTP so that everyone can check it. Since then, I admitted several times that the weighting is off and even proposed a different one, where Aligulac and match win rates only make up 5% each. Also I never propped Serral lol. He gets penalized the most by subjective decisions I had to make (including non-prime years in tournament win percentage and leave out the inflation-bonus in the match win rate). I remember being asked in the comments of my first article, if I would post the results, if changing the methodology of the original article would mean that Serral would not come out on top in any of the metrics. Well, as it is obvious now, I have no issue accepting that Serral is not the best in tournament win percentage. But it seems like people are trying to critique my result without actually addressing any core methodological feature (except for weighting, which was explained weeks ago). What I would find more troubling than obvious weighting issues is a GOAT ranking that presents itself as statistical but in the end is a post-hoc feeling based scripture. And of course you can make a ranking that is a lot more objective than someone's gut feeling. Not being in alignment with every subjective feeling for era-difference or tournament distinctions does not mean that going as close as possible in a statistical evaluation is the same as a random StarCraft II player that thinks this or that player is the GOAT. We can very clearly see from the statistical evaluation that Serral is way ahead of Maru and Rogue in the time-frame that they played together. We can see that Life was ahead of Rain. We can also see what everyone subjectively thought for a long time: That Rogue has pretty mixed results. Plus, subjectivity can be put into a bare minimum. 1. We need a point/ranking system for the tournaments (similar to what Miz and I did) 2. We need some kind of era-multiplier for different time brackets in terms of how hard it was to win (This can also be calculated more or less ok-ish as well) OR come to the conclusion that all time periods were the same, which I doubt is realistic or correct 3. Decide which metrics are the most important and then add a weighting (the community here and on reddit already agree that tournament score and tournament win participation are the most important These are all the subjective things one needs and the community will probably come to a close proximity. But, even in the case that we have heavy disagreements here: When I revise my method one more time, I will try to make it in a way that people can tell me their desired ratios and I can post the result based on these suggestions. When all my excels are interlinked, I can easily take community suggestions for these three and we can look at how hard we would need to torture the data to make this or that player come out on top. You do not penalize Serral, though I think we all understand rhetorically why you say you do. You give him bonuses. A lot of them. And you only admit it when people call out the more obvious and ridiculous ways in which you do it like your 20% Aligulac weighting. The basic logical error you make is that you are calling all your criteria modifiers "penalties" to Serral but whether they are penalties or bonuses entirely depends on baselines versus expectations in your modifiers. Take your tournament win % criteria. Again you call your 20% (or whatever it was) modifier a "penalty" to Serral. But the question is whether that is too high or too low or just right given that Serral competed exclusively in tournaments that are significantly easier to win than KILs, which have dozens and sometimes hundreds of additional competitors, harder groups stages, more group stages, etc. If it's just right, there's no penalty or bonus to Serral, if it's too high, it's a penalty to Serral, and if it's too low, it's a bonus to Serral. I simply do not agree that the 20% discount reflects the comparative odds of winning a 2020 Dreamhack Masters versus a 2013 GSL or SSL. It's probably 5X easier to win the former than the latter imho. In this specific example, though, it doesn't even matter because it's actually a bonus to Serral irrespective of whether the 20% modifier is right or wrong as he is the only player on your GOAT list that does not have his tournament win % diluted by participation in the signiifcantly-harder-to-win KILs. This is only one example of how you claim to be penalizing Serral when you're actually giving him bonuses in your model. This kind of logical error permeates your model. At first I was giving you the benefit of the doubt that you did not understand some of these concepts in predictive modeling, but it's pretty clear that you do and are just hoping that others do not understand so you can get away with saying over and over again that you are penalizing Serral. Your entire goal is to say that no matter how much you penalize Serral, he is the GOAT, and because you're so focused on this, it's producing many errors and requiring you to go to great lengths to obscure what it is you are doing. The simple fact is that your model giving Serral 2X the GOAT points of the next highest GOAT and 5X the GOAT points of Rogue is on its face not credible. You try to get around this by hiding behind your methodology, but in just a very cursory review of your methodology like I've done here reveals huge errors, almost all of which are designed to give Serral bonus GOAT points. Wouldn't it be better at this point just to admin you've created a Serral = GOAT calculator? It's totally legitimate to call Serral the GOAT--he's probably most people's GOAT, even most players probably think he's the GOAT. You created a highly subjective model to crown Serral the GOAT. It's no different than this GOAT list, just different subjective criteria. It is not rhetorical to say that I penalized Serral, because I did: 1a. Do you deny that the Koreans played more lower tier players in comparison to Serral? Or do you deny that I wrote that I did not include this correction, which is detrimental to Serral? 1b. Do you deny that including non-prime years, which I did, penalizes/hurts Serral the most according to the data? 2. In the article - as well as multiple times in the other thread (and even in the very post you commented on and directly to you in the other thread!!) I explained that this weighting was proposed by ChatGTP. I still don't understand why you keep pounding on that, tbh. As WombaT wrote out in the other thread, the weighting is entirely superfluous. The most important revelation was that most players have mixed results, while Serral is dominating them all. You make it seem like the weighting was there to push Serral, which is not the case. You have no proof for this accusation. On the contrary, I posted the exact text that I used to let ChatGTP decide on the weighting, to let you guys double check. I only did so because I knew based on the results, that the weighting does not change in the first place and I couldn’t decide on a specific weighting, which I explained ad nauseam in the article as well as in the comments. First place would only change with an insane weighting-setup for efficiency… then Life (who most people disregard anyway) would be on top, with Serral being 2nd. Arguing with the weighting result is utterly pointless… As this is your (and other critics) biggest point which you all come back to every time, I doubt that is much more substance to gain from you guys. 3a. In the tournament win %, similar to most metrics, I discounted all region-locked tournaments. This is a penalty to Serral, although a needed one to deny his easier tournament wins. But still it is a penalty; especially to deny them entirely, as Korean pros had rather easy tournament wins as well. 3b. You might disagree that the bonus is not enough according to your subjective feeling. My reasoning for using 20%: I explained in detail why a larger player pool, especially in mostly tier 2, 3 and 4 players, would not decrease a tournament win percentage that much. - doubling the player pool would most likely mean just one extra round. If that round was a group stage, it would naturally mean that you’d have to play more lower tier player statistically, boosting your match win rate - I reasoned that this fact means that average place would need a higher era-multiplier than tournament win percentage - Now looking at Serral’s average win rate of around 82% in his years from 2018-2024 subtracted with 2% (some lower match win for era but also considering that his win rates are deflated versus the Koreans because of 1a, also taking into account that earlier group stages are easier to win than knockout brackets), these more rounds would mean a win probability of roughly 80%. As an example, going from 4 rounds to 5 rounds would statistically mean a win probability 40% to 33%, meaning a loss of 17,5% or a boost of 21,2%. - Now, comparing DH 2020 to a GSL or SSL 2013: Not all tournaments that other players have won were a GSL though. What about the 2013 Hot6ix Cup or ASUS ROG NorthCon? And didn’t Serral win uberly stacked tournaments as well? You are committing a false equivalence/improper comparison fallacy here. In essence, you are right that one can’t compare a GSL to a DH. But that is also not what I did. There were not only GSLs in 2013 (as a matter of fact, there was only one) and Serral won more stacked tournaments as well. It also borders on a strawman, as my approach was not designed to precisely reflect the odds between only one of the hardest 2013 tournaments vs a specific 2020 event, but rather roughly corrects across periods. - Do you further realize that arguing like this from an era/period-perspective also argues against Rogue’s and Maru’s case? - What adjustment would you find appropriate? Overall it is rather the modern Koreans that benefitted in this metric, as Serral, Clem and mostly Reynor were missing in the GSLs that they won. Of course Serral wouldn’t have the same tournament win percentage had he been a full-time pro at his peak playing in the prime era. But the win percentages of INno, Rain, etc. would drop as well, had Mvp, Serral and Clem played there. I explained this already in the article. “Again you call your 20% (or whatever it was) modifier a "penalty" to Serral” Can you give the exact quote? This would be much more fruitful if you didn’t try to rely on notions like my method was not statistical, keep using arguments that have been addressed or straight out deny that Serral was penalized on several occasions, which he clearly was. Talk about the points ratio in the tournaments or the era-multiplier or your preferred weighting. Or - as you by your own account don’t care for statistical GOAT debates anyway - simply stay out of them. But making unsubstantiated accusations like me making a Serral GOAT-calculator or repeating the same already addressed notions is absolutely ridiculous. You "penalize" Serral within metrics that heavily favor him. For comparison, it's like if I would incorporate a metric for number of seperate years with a playoffs finish (a metric obviously favoring Maru since he had the longest career), giving it 25% of the overall score, and then acting like I penalized Maru because I considered korean tournaments the same as foreigner-only tournaments, despite being obviously harder. I already said in the article that there are a million ways to fragment the data... one could also make a scoreboard for best cannon rush defender or what you just suggested. But it simply makes no sense, as I used data that covers the entirety of the game and metrics that are most valuable to determine the Greatest of All Time. If you disagree, simply explain why your example is a worthy metric. But your comparison is completely off, as the tournament win percentage is a simple data set. Nothing of it has been modified like you did. The only thing that was modified was - disregarding region locks, which is a (needed) penalty for Serral - add a multiplier that boosted everything pre-2018 and penalized everything post-2018 (which penalized Serral among others) So which metric does favor Serral? And why exactly does it "favor" him?
If it's entirely reasonable to modify the data to enable comparability, I wouldn't call it a "penalty".
If we're talking about tournament wins for example, the raw data set should include online cups, however as we both agree it makes no sense to include them in a GOAT analysis and exclude them.
Despite that I wouldn't call it a Clem "penalty", it just removes data which is not relevant for the topic. That's how I feel about your so called Serral "penalty".
|
On July 15 2025 20:54 Charoisaur wrote:Show nested quote +On July 15 2025 20:20 PremoBeats wrote:On July 15 2025 20:14 MJG wrote:Premo should take their own advice: On August 10 2024 00:10 PremoBeats wrote:
Seriously.. if there is someone out there taking this obvious bait then it is on them... I don't think Charoisaur and rwala are baiting me from the way they reply... they truly think I favor Serral imo. But perhaps that just went past me, lol. On July 15 2025 19:57 Charoisaur wrote:On July 15 2025 19:31 PremoBeats wrote:On July 15 2025 17:32 Charoisaur wrote:On July 15 2025 03:59 rwala wrote:On July 13 2025 21:29 PremoBeats wrote:On July 13 2025 05:44 Charoisaur wrote:On July 12 2025 06:50 Balnazza wrote: but at the same time, it doesn't wank itself into a corner by bias after bias This is literally what you did though? Just to take the WCS eu/am again, it completly fulfills your criteria and yet you disregard it. Yes, the qualifier was open to ALL koreans, they just had to move to Europe or America. That is essentially the same requirement GSL had, you had to live in Korea for a while. The biggest reason not more koreans did it was most likely Proleague, because you couldn't really play Proleague when you played WCS. But that isn't an excluding factor. This was pre-regionlock, when koreans were actually forced to be residents of Europe or America. But the whole criticism before that was how lackluster some koreans were in "their" new region, how little they contributed to the regional scenes. If you want to take prizemoney, just take prizemoney. And yes, then you will have to live with the fact that Clem (by winning EWC), Reynor (by winning Gamers8) and even Maru to a degree (by winning WESG) have a bit of inflated stats, but that is the hand you have been dealt with if you pick prizemoney as the defining statistic. But the second you introduced your "fine-tuning", you made yourself vulnerable...and as you notice, a lot of people (if not all of them) don't agree with your fine-tuning, not even the ones that agree with the outcome. Because it is such a biased and random bunch of multipliers. Just as an example I will point out again the HSC I mentioned earlier, that had barely anyone of notice in attendance but gave out more points than Katowice the next year, just because both tournaments were in different "eras"...divided by two months. damn did the scene drop in two months! And don't get me started on the highly subjective "imbalance"-multiplier... No GOAT list will ever be perfect, that's impossible. If we had like 40 years of SC2 that would have been somewhat the same we would have a good metric to decide by, but considering how much shift and change there has been in the ~15 years lifespan of this game, it will always come down to what statistic you highlight and what metric you choose. But sadly, your way just wasn't solid at all. It was highly biased and too radical to correctly adjust for change (not that you particularly need to). But the funniest part is honestly that you think your approach is completly unbiased... Yeah, this measurement is quite faulty I think, but it brings a great point across which ironically is the one OP quoted in his post "If you torture the data long enough, it will confess to anything" At the end of the day you can use all the fancy stats and data you want, but it doesn't make your ranking any less subjective than someone's guts feeling, whether it's premobeats way overvaluing things like winrates or aligulac rating to prob Serral or the thing ejozl did here. And we all know the one true Goat ranking is this one: 1. Serral 2. Maru (extremely close to Serral) 3. Inno 4. Rogue 5. Life 6. Zest 7. sOs 8. herO 9. Dark 10. Stats I think that quote was put there unironically to highlight the sarcastic over-the-top nature of the article. But why do you mischaracterize what was happening? I wrote in the article that I did not come up with this weighting. I even posted the exact quote that I inserted into ChatGTP so that everyone can check it. Since then, I admitted several times that the weighting is off and even proposed a different one, where Aligulac and match win rates only make up 5% each. Also I never propped Serral lol. He gets penalized the most by subjective decisions I had to make (including non-prime years in tournament win percentage and leave out the inflation-bonus in the match win rate). I remember being asked in the comments of my first article, if I would post the results, if changing the methodology of the original article would mean that Serral would not come out on top in any of the metrics. Well, as it is obvious now, I have no issue accepting that Serral is not the best in tournament win percentage. But it seems like people are trying to critique my result without actually addressing any core methodological feature (except for weighting, which was explained weeks ago). What I would find more troubling than obvious weighting issues is a GOAT ranking that presents itself as statistical but in the end is a post-hoc feeling based scripture. And of course you can make a ranking that is a lot more objective than someone's gut feeling. Not being in alignment with every subjective feeling for era-difference or tournament distinctions does not mean that going as close as possible in a statistical evaluation is the same as a random StarCraft II player that thinks this or that player is the GOAT. We can very clearly see from the statistical evaluation that Serral is way ahead of Maru and Rogue in the time-frame that they played together. We can see that Life was ahead of Rain. We can also see what everyone subjectively thought for a long time: That Rogue has pretty mixed results. Plus, subjectivity can be put into a bare minimum. 1. We need a point/ranking system for the tournaments (similar to what Miz and I did) 2. We need some kind of era-multiplier for different time brackets in terms of how hard it was to win (This can also be calculated more or less ok-ish as well) OR come to the conclusion that all time periods were the same, which I doubt is realistic or correct 3. Decide which metrics are the most important and then add a weighting (the community here and on reddit already agree that tournament score and tournament win participation are the most important These are all the subjective things one needs and the community will probably come to a close proximity. But, even in the case that we have heavy disagreements here: When I revise my method one more time, I will try to make it in a way that people can tell me their desired ratios and I can post the result based on these suggestions. When all my excels are interlinked, I can easily take community suggestions for these three and we can look at how hard we would need to torture the data to make this or that player come out on top. You do not penalize Serral, though I think we all understand rhetorically why you say you do. You give him bonuses. A lot of them. And you only admit it when people call out the more obvious and ridiculous ways in which you do it like your 20% Aligulac weighting. The basic logical error you make is that you are calling all your criteria modifiers "penalties" to Serral but whether they are penalties or bonuses entirely depends on baselines versus expectations in your modifiers. Take your tournament win % criteria. Again you call your 20% (or whatever it was) modifier a "penalty" to Serral. But the question is whether that is too high or too low or just right given that Serral competed exclusively in tournaments that are significantly easier to win than KILs, which have dozens and sometimes hundreds of additional competitors, harder groups stages, more group stages, etc. If it's just right, there's no penalty or bonus to Serral, if it's too high, it's a penalty to Serral, and if it's too low, it's a bonus to Serral. I simply do not agree that the 20% discount reflects the comparative odds of winning a 2020 Dreamhack Masters versus a 2013 GSL or SSL. It's probably 5X easier to win the former than the latter imho. In this specific example, though, it doesn't even matter because it's actually a bonus to Serral irrespective of whether the 20% modifier is right or wrong as he is the only player on your GOAT list that does not have his tournament win % diluted by participation in the signiifcantly-harder-to-win KILs. This is only one example of how you claim to be penalizing Serral when you're actually giving him bonuses in your model. This kind of logical error permeates your model. At first I was giving you the benefit of the doubt that you did not understand some of these concepts in predictive modeling, but it's pretty clear that you do and are just hoping that others do not understand so you can get away with saying over and over again that you are penalizing Serral. Your entire goal is to say that no matter how much you penalize Serral, he is the GOAT, and because you're so focused on this, it's producing many errors and requiring you to go to great lengths to obscure what it is you are doing. The simple fact is that your model giving Serral 2X the GOAT points of the next highest GOAT and 5X the GOAT points of Rogue is on its face not credible. You try to get around this by hiding behind your methodology, but in just a very cursory review of your methodology like I've done here reveals huge errors, almost all of which are designed to give Serral bonus GOAT points. Wouldn't it be better at this point just to admin you've created a Serral = GOAT calculator? It's totally legitimate to call Serral the GOAT--he's probably most people's GOAT, even most players probably think he's the GOAT. You created a highly subjective model to crown Serral the GOAT. It's no different than this GOAT list, just different subjective criteria. Well said rwala's reply consists mostly of straw men, false comparisons and already addressed arguments... what exactly do you think is well said about it? On July 15 2025 03:59 rwala wrote:On July 13 2025 21:29 PremoBeats wrote:On July 13 2025 05:44 Charoisaur wrote:On July 12 2025 06:50 Balnazza wrote: but at the same time, it doesn't wank itself into a corner by bias after bias This is literally what you did though? Just to take the WCS eu/am again, it completly fulfills your criteria and yet you disregard it. Yes, the qualifier was open to ALL koreans, they just had to move to Europe or America. That is essentially the same requirement GSL had, you had to live in Korea for a while. The biggest reason not more koreans did it was most likely Proleague, because you couldn't really play Proleague when you played WCS. But that isn't an excluding factor. This was pre-regionlock, when koreans were actually forced to be residents of Europe or America. But the whole criticism before that was how lackluster some koreans were in "their" new region, how little they contributed to the regional scenes. If you want to take prizemoney, just take prizemoney. And yes, then you will have to live with the fact that Clem (by winning EWC), Reynor (by winning Gamers8) and even Maru to a degree (by winning WESG) have a bit of inflated stats, but that is the hand you have been dealt with if you pick prizemoney as the defining statistic. But the second you introduced your "fine-tuning", you made yourself vulnerable...and as you notice, a lot of people (if not all of them) don't agree with your fine-tuning, not even the ones that agree with the outcome. Because it is such a biased and random bunch of multipliers. Just as an example I will point out again the HSC I mentioned earlier, that had barely anyone of notice in attendance but gave out more points than Katowice the next year, just because both tournaments were in different "eras"...divided by two months. damn did the scene drop in two months! And don't get me started on the highly subjective "imbalance"-multiplier... No GOAT list will ever be perfect, that's impossible. If we had like 40 years of SC2 that would have been somewhat the same we would have a good metric to decide by, but considering how much shift and change there has been in the ~15 years lifespan of this game, it will always come down to what statistic you highlight and what metric you choose. But sadly, your way just wasn't solid at all. It was highly biased and too radical to correctly adjust for change (not that you particularly need to). But the funniest part is honestly that you think your approach is completly unbiased... Yeah, this measurement is quite faulty I think, but it brings a great point across which ironically is the one OP quoted in his post "If you torture the data long enough, it will confess to anything" At the end of the day you can use all the fancy stats and data you want, but it doesn't make your ranking any less subjective than someone's guts feeling, whether it's premobeats way overvaluing things like winrates or aligulac rating to prob Serral or the thing ejozl did here. And we all know the one true Goat ranking is this one: 1. Serral 2. Maru (extremely close to Serral) 3. Inno 4. Rogue 5. Life 6. Zest 7. sOs 8. herO 9. Dark 10. Stats I think that quote was put there unironically to highlight the sarcastic over-the-top nature of the article. But why do you mischaracterize what was happening? I wrote in the article that I did not come up with this weighting. I even posted the exact quote that I inserted into ChatGTP so that everyone can check it. Since then, I admitted several times that the weighting is off and even proposed a different one, where Aligulac and match win rates only make up 5% each. Also I never propped Serral lol. He gets penalized the most by subjective decisions I had to make (including non-prime years in tournament win percentage and leave out the inflation-bonus in the match win rate). I remember being asked in the comments of my first article, if I would post the results, if changing the methodology of the original article would mean that Serral would not come out on top in any of the metrics. Well, as it is obvious now, I have no issue accepting that Serral is not the best in tournament win percentage. But it seems like people are trying to critique my result without actually addressing any core methodological feature (except for weighting, which was explained weeks ago). What I would find more troubling than obvious weighting issues is a GOAT ranking that presents itself as statistical but in the end is a post-hoc feeling based scripture. And of course you can make a ranking that is a lot more objective than someone's gut feeling. Not being in alignment with every subjective feeling for era-difference or tournament distinctions does not mean that going as close as possible in a statistical evaluation is the same as a random StarCraft II player that thinks this or that player is the GOAT. We can very clearly see from the statistical evaluation that Serral is way ahead of Maru and Rogue in the time-frame that they played together. We can see that Life was ahead of Rain. We can also see what everyone subjectively thought for a long time: That Rogue has pretty mixed results. Plus, subjectivity can be put into a bare minimum. 1. We need a point/ranking system for the tournaments (similar to what Miz and I did) 2. We need some kind of era-multiplier for different time brackets in terms of how hard it was to win (This can also be calculated more or less ok-ish as well) OR come to the conclusion that all time periods were the same, which I doubt is realistic or correct 3. Decide which metrics are the most important and then add a weighting (the community here and on reddit already agree that tournament score and tournament win participation are the most important These are all the subjective things one needs and the community will probably come to a close proximity. But, even in the case that we have heavy disagreements here: When I revise my method one more time, I will try to make it in a way that people can tell me their desired ratios and I can post the result based on these suggestions. When all my excels are interlinked, I can easily take community suggestions for these three and we can look at how hard we would need to torture the data to make this or that player come out on top. You do not penalize Serral, though I think we all understand rhetorically why you say you do. You give him bonuses. A lot of them. And you only admit it when people call out the more obvious and ridiculous ways in which you do it like your 20% Aligulac weighting. The basic logical error you make is that you are calling all your criteria modifiers "penalties" to Serral but whether they are penalties or bonuses entirely depends on baselines versus expectations in your modifiers. Take your tournament win % criteria. Again you call your 20% (or whatever it was) modifier a "penalty" to Serral. But the question is whether that is too high or too low or just right given that Serral competed exclusively in tournaments that are significantly easier to win than KILs, which have dozens and sometimes hundreds of additional competitors, harder groups stages, more group stages, etc. If it's just right, there's no penalty or bonus to Serral, if it's too high, it's a penalty to Serral, and if it's too low, it's a bonus to Serral. I simply do not agree that the 20% discount reflects the comparative odds of winning a 2020 Dreamhack Masters versus a 2013 GSL or SSL. It's probably 5X easier to win the former than the latter imho. In this specific example, though, it doesn't even matter because it's actually a bonus to Serral irrespective of whether the 20% modifier is right or wrong as he is the only player on your GOAT list that does not have his tournament win % diluted by participation in the signiifcantly-harder-to-win KILs. This is only one example of how you claim to be penalizing Serral when you're actually giving him bonuses in your model. This kind of logical error permeates your model. At first I was giving you the benefit of the doubt that you did not understand some of these concepts in predictive modeling, but it's pretty clear that you do and are just hoping that others do not understand so you can get away with saying over and over again that you are penalizing Serral. Your entire goal is to say that no matter how much you penalize Serral, he is the GOAT, and because you're so focused on this, it's producing many errors and requiring you to go to great lengths to obscure what it is you are doing. The simple fact is that your model giving Serral 2X the GOAT points of the next highest GOAT and 5X the GOAT points of Rogue is on its face not credible. You try to get around this by hiding behind your methodology, but in just a very cursory review of your methodology like I've done here reveals huge errors, almost all of which are designed to give Serral bonus GOAT points. Wouldn't it be better at this point just to admin you've created a Serral = GOAT calculator? It's totally legitimate to call Serral the GOAT--he's probably most people's GOAT, even most players probably think he's the GOAT. You created a highly subjective model to crown Serral the GOAT. It's no different than this GOAT list, just different subjective criteria. It is not rhetorical to say that I penalized Serral, because I did: 1a. Do you deny that the Koreans played more lower tier players in comparison to Serral? Or do you deny that I wrote that I did not include this correction, which is detrimental to Serral? 1b. Do you deny that including non-prime years, which I did, penalizes/hurts Serral the most according to the data? 2. In the article - as well as multiple times in the other thread (and even in the very post you commented on and directly to you in the other thread!!) I explained that this weighting was proposed by ChatGTP. I still don't understand why you keep pounding on that, tbh. As WombaT wrote out in the other thread, the weighting is entirely superfluous. The most important revelation was that most players have mixed results, while Serral is dominating them all. You make it seem like the weighting was there to push Serral, which is not the case. You have no proof for this accusation. On the contrary, I posted the exact text that I used to let ChatGTP decide on the weighting, to let you guys double check. I only did so because I knew based on the results, that the weighting does not change in the first place and I couldn’t decide on a specific weighting, which I explained ad nauseam in the article as well as in the comments. First place would only change with an insane weighting-setup for efficiency… then Life (who most people disregard anyway) would be on top, with Serral being 2nd. Arguing with the weighting result is utterly pointless… As this is your (and other critics) biggest point which you all come back to every time, I doubt that is much more substance to gain from you guys. 3a. In the tournament win %, similar to most metrics, I discounted all region-locked tournaments. This is a penalty to Serral, although a needed one to deny his easier tournament wins. But still it is a penalty; especially to deny them entirely, as Korean pros had rather easy tournament wins as well. 3b. You might disagree that the bonus is not enough according to your subjective feeling. My reasoning for using 20%: I explained in detail why a larger player pool, especially in mostly tier 2, 3 and 4 players, would not decrease a tournament win percentage that much. - doubling the player pool would most likely mean just one extra round. If that round was a group stage, it would naturally mean that you’d have to play more lower tier player statistically, boosting your match win rate - I reasoned that this fact means that average place would need a higher era-multiplier than tournament win percentage - Now looking at Serral’s average win rate of around 82% in his years from 2018-2024 subtracted with 2% (some lower match win for era but also considering that his win rates are deflated versus the Koreans because of 1a, also taking into account that earlier group stages are easier to win than knockout brackets), these more rounds would mean a win probability of roughly 80%. As an example, going from 4 rounds to 5 rounds would statistically mean a win probability 40% to 33%, meaning a loss of 17,5% or a boost of 21,2%. - Now, comparing DH 2020 to a GSL or SSL 2013: Not all tournaments that other players have won were a GSL though. What about the 2013 Hot6ix Cup or ASUS ROG NorthCon? And didn’t Serral win uberly stacked tournaments as well? You are committing a false equivalence/improper comparison fallacy here. In essence, you are right that one can’t compare a GSL to a DH. But that is also not what I did. There were not only GSLs in 2013 (as a matter of fact, there was only one) and Serral won more stacked tournaments as well. It also borders on a strawman, as my approach was not designed to precisely reflect the odds between only one of the hardest 2013 tournaments vs a specific 2020 event, but rather roughly corrects across periods. - Do you further realize that arguing like this from an era/period-perspective also argues against Rogue’s and Maru’s case? - What adjustment would you find appropriate? Overall it is rather the modern Koreans that benefitted in this metric, as Serral, Clem and mostly Reynor were missing in the GSLs that they won. Of course Serral wouldn’t have the same tournament win percentage had he been a full-time pro at his peak playing in the prime era. But the win percentages of INno, Rain, etc. would drop as well, had Mvp, Serral and Clem played there. I explained this already in the article. “Again you call your 20% (or whatever it was) modifier a "penalty" to Serral” Can you give the exact quote? This would be much more fruitful if you didn’t try to rely on notions like my method was not statistical, keep using arguments that have been addressed or straight out deny that Serral was penalized on several occasions, which he clearly was. Talk about the points ratio in the tournaments or the era-multiplier or your preferred weighting. Or - as you by your own account don’t care for statistical GOAT debates anyway - simply stay out of them. But making unsubstantiated accusations like me making a Serral GOAT-calculator or repeating the same already addressed notions is absolutely ridiculous. You "penalize" Serral within metrics that heavily favor him. For comparison, it's like if I would incorporate a metric for number of seperate years with a playoffs finish (a metric obviously favoring Maru since he had the longest career), giving it 25% of the overall score, and then acting like I penalized Maru because I considered korean tournaments the same as foreigner-only tournaments, despite being obviously harder. Which metric does favor Serral? And why does it "favor" him? Because he is good in it? Aligulac rating, match winrate, tournament winrate, average placing are all metrics that favor Serral. Yes, because he's good at it, but you attributed all those metrics an unreasonably high influence on the final score, especially because some of them measure very similar things and thus have massive collinearity e.g. aligulac rating and average placement. Metrics in which Serral wouldn't perform as well however are excluded, like longevity focused metrics, e.g. the aforementioned number of seperate years with a top x finish.
For the last time: The unreasonably high influence was not suggested by me.. and it would not have mattered, because Serral placed 1st in each and every one of them, except the least important one. Is this so hard to understand?
Did you forget that I said a dozen times that the weighting was off and already suggested a different one?
Longevity is measued in the tournament score, which was and will be weighted the heaviest.
Aligulac is also measuring longevity... the longer you stay on top, the more points you have... it is calculated against the entirety of the game.
Yes, your metric is fragmenting unncessarily, which none of mine do. I looked at the whole careers and only pointed out great results based on years... but they were not included to form the final score.
By fragmenting careers into multiple "peak segments" or time blocks you overemphasize people who had peaks. That is why a mix of longevity and efficiency in my opinion is the best. Otherwise you discredit players who were able to achieve top peaks for most of their career. Your suggestion is an artificial inflation for no reason, as your metric is present in "overarching" metrics anyway, as it shows dominance. Looking at the top 4 finishes basically is the tournament score.
On July 15 2025 21:03 Charoisaur wrote:Show nested quote +On July 15 2025 20:20 PremoBeats wrote:On July 15 2025 20:14 MJG wrote:Premo should take their own advice: On August 10 2024 00:10 PremoBeats wrote:
Seriously.. if there is someone out there taking this obvious bait then it is on them... I don't think Charoisaur and rwala are baiting me from the way they reply... they truly think I favor Serral imo. But perhaps that just went past me, lol. On July 15 2025 19:57 Charoisaur wrote:On July 15 2025 19:31 PremoBeats wrote:On July 15 2025 17:32 Charoisaur wrote:On July 15 2025 03:59 rwala wrote:On July 13 2025 21:29 PremoBeats wrote:On July 13 2025 05:44 Charoisaur wrote:On July 12 2025 06:50 Balnazza wrote: but at the same time, it doesn't wank itself into a corner by bias after bias This is literally what you did though? Just to take the WCS eu/am again, it completly fulfills your criteria and yet you disregard it. Yes, the qualifier was open to ALL koreans, they just had to move to Europe or America. That is essentially the same requirement GSL had, you had to live in Korea for a while. The biggest reason not more koreans did it was most likely Proleague, because you couldn't really play Proleague when you played WCS. But that isn't an excluding factor. This was pre-regionlock, when koreans were actually forced to be residents of Europe or America. But the whole criticism before that was how lackluster some koreans were in "their" new region, how little they contributed to the regional scenes. If you want to take prizemoney, just take prizemoney. And yes, then you will have to live with the fact that Clem (by winning EWC), Reynor (by winning Gamers8) and even Maru to a degree (by winning WESG) have a bit of inflated stats, but that is the hand you have been dealt with if you pick prizemoney as the defining statistic. But the second you introduced your "fine-tuning", you made yourself vulnerable...and as you notice, a lot of people (if not all of them) don't agree with your fine-tuning, not even the ones that agree with the outcome. Because it is such a biased and random bunch of multipliers. Just as an example I will point out again the HSC I mentioned earlier, that had barely anyone of notice in attendance but gave out more points than Katowice the next year, just because both tournaments were in different "eras"...divided by two months. damn did the scene drop in two months! And don't get me started on the highly subjective "imbalance"-multiplier... No GOAT list will ever be perfect, that's impossible. If we had like 40 years of SC2 that would have been somewhat the same we would have a good metric to decide by, but considering how much shift and change there has been in the ~15 years lifespan of this game, it will always come down to what statistic you highlight and what metric you choose. But sadly, your way just wasn't solid at all. It was highly biased and too radical to correctly adjust for change (not that you particularly need to). But the funniest part is honestly that you think your approach is completly unbiased... Yeah, this measurement is quite faulty I think, but it brings a great point across which ironically is the one OP quoted in his post "If you torture the data long enough, it will confess to anything" At the end of the day you can use all the fancy stats and data you want, but it doesn't make your ranking any less subjective than someone's guts feeling, whether it's premobeats way overvaluing things like winrates or aligulac rating to prob Serral or the thing ejozl did here. And we all know the one true Goat ranking is this one: 1. Serral 2. Maru (extremely close to Serral) 3. Inno 4. Rogue 5. Life 6. Zest 7. sOs 8. herO 9. Dark 10. Stats I think that quote was put there unironically to highlight the sarcastic over-the-top nature of the article. But why do you mischaracterize what was happening? I wrote in the article that I did not come up with this weighting. I even posted the exact quote that I inserted into ChatGTP so that everyone can check it. Since then, I admitted several times that the weighting is off and even proposed a different one, where Aligulac and match win rates only make up 5% each. Also I never propped Serral lol. He gets penalized the most by subjective decisions I had to make (including non-prime years in tournament win percentage and leave out the inflation-bonus in the match win rate). I remember being asked in the comments of my first article, if I would post the results, if changing the methodology of the original article would mean that Serral would not come out on top in any of the metrics. Well, as it is obvious now, I have no issue accepting that Serral is not the best in tournament win percentage. But it seems like people are trying to critique my result without actually addressing any core methodological feature (except for weighting, which was explained weeks ago). What I would find more troubling than obvious weighting issues is a GOAT ranking that presents itself as statistical but in the end is a post-hoc feeling based scripture. And of course you can make a ranking that is a lot more objective than someone's gut feeling. Not being in alignment with every subjective feeling for era-difference or tournament distinctions does not mean that going as close as possible in a statistical evaluation is the same as a random StarCraft II player that thinks this or that player is the GOAT. We can very clearly see from the statistical evaluation that Serral is way ahead of Maru and Rogue in the time-frame that they played together. We can see that Life was ahead of Rain. We can also see what everyone subjectively thought for a long time: That Rogue has pretty mixed results. Plus, subjectivity can be put into a bare minimum. 1. We need a point/ranking system for the tournaments (similar to what Miz and I did) 2. We need some kind of era-multiplier for different time brackets in terms of how hard it was to win (This can also be calculated more or less ok-ish as well) OR come to the conclusion that all time periods were the same, which I doubt is realistic or correct 3. Decide which metrics are the most important and then add a weighting (the community here and on reddit already agree that tournament score and tournament win participation are the most important These are all the subjective things one needs and the community will probably come to a close proximity. But, even in the case that we have heavy disagreements here: When I revise my method one more time, I will try to make it in a way that people can tell me their desired ratios and I can post the result based on these suggestions. When all my excels are interlinked, I can easily take community suggestions for these three and we can look at how hard we would need to torture the data to make this or that player come out on top. You do not penalize Serral, though I think we all understand rhetorically why you say you do. You give him bonuses. A lot of them. And you only admit it when people call out the more obvious and ridiculous ways in which you do it like your 20% Aligulac weighting. The basic logical error you make is that you are calling all your criteria modifiers "penalties" to Serral but whether they are penalties or bonuses entirely depends on baselines versus expectations in your modifiers. Take your tournament win % criteria. Again you call your 20% (or whatever it was) modifier a "penalty" to Serral. But the question is whether that is too high or too low or just right given that Serral competed exclusively in tournaments that are significantly easier to win than KILs, which have dozens and sometimes hundreds of additional competitors, harder groups stages, more group stages, etc. If it's just right, there's no penalty or bonus to Serral, if it's too high, it's a penalty to Serral, and if it's too low, it's a bonus to Serral. I simply do not agree that the 20% discount reflects the comparative odds of winning a 2020 Dreamhack Masters versus a 2013 GSL or SSL. It's probably 5X easier to win the former than the latter imho. In this specific example, though, it doesn't even matter because it's actually a bonus to Serral irrespective of whether the 20% modifier is right or wrong as he is the only player on your GOAT list that does not have his tournament win % diluted by participation in the signiifcantly-harder-to-win KILs. This is only one example of how you claim to be penalizing Serral when you're actually giving him bonuses in your model. This kind of logical error permeates your model. At first I was giving you the benefit of the doubt that you did not understand some of these concepts in predictive modeling, but it's pretty clear that you do and are just hoping that others do not understand so you can get away with saying over and over again that you are penalizing Serral. Your entire goal is to say that no matter how much you penalize Serral, he is the GOAT, and because you're so focused on this, it's producing many errors and requiring you to go to great lengths to obscure what it is you are doing. The simple fact is that your model giving Serral 2X the GOAT points of the next highest GOAT and 5X the GOAT points of Rogue is on its face not credible. You try to get around this by hiding behind your methodology, but in just a very cursory review of your methodology like I've done here reveals huge errors, almost all of which are designed to give Serral bonus GOAT points. Wouldn't it be better at this point just to admin you've created a Serral = GOAT calculator? It's totally legitimate to call Serral the GOAT--he's probably most people's GOAT, even most players probably think he's the GOAT. You created a highly subjective model to crown Serral the GOAT. It's no different than this GOAT list, just different subjective criteria. Well said rwala's reply consists mostly of straw men, false comparisons and already addressed arguments... what exactly do you think is well said about it? On July 15 2025 03:59 rwala wrote:On July 13 2025 21:29 PremoBeats wrote:On July 13 2025 05:44 Charoisaur wrote:On July 12 2025 06:50 Balnazza wrote: but at the same time, it doesn't wank itself into a corner by bias after bias This is literally what you did though? Just to take the WCS eu/am again, it completly fulfills your criteria and yet you disregard it. Yes, the qualifier was open to ALL koreans, they just had to move to Europe or America. That is essentially the same requirement GSL had, you had to live in Korea for a while. The biggest reason not more koreans did it was most likely Proleague, because you couldn't really play Proleague when you played WCS. But that isn't an excluding factor. This was pre-regionlock, when koreans were actually forced to be residents of Europe or America. But the whole criticism before that was how lackluster some koreans were in "their" new region, how little they contributed to the regional scenes. If you want to take prizemoney, just take prizemoney. And yes, then you will have to live with the fact that Clem (by winning EWC), Reynor (by winning Gamers8) and even Maru to a degree (by winning WESG) have a bit of inflated stats, but that is the hand you have been dealt with if you pick prizemoney as the defining statistic. But the second you introduced your "fine-tuning", you made yourself vulnerable...and as you notice, a lot of people (if not all of them) don't agree with your fine-tuning, not even the ones that agree with the outcome. Because it is such a biased and random bunch of multipliers. Just as an example I will point out again the HSC I mentioned earlier, that had barely anyone of notice in attendance but gave out more points than Katowice the next year, just because both tournaments were in different "eras"...divided by two months. damn did the scene drop in two months! And don't get me started on the highly subjective "imbalance"-multiplier... No GOAT list will ever be perfect, that's impossible. If we had like 40 years of SC2 that would have been somewhat the same we would have a good metric to decide by, but considering how much shift and change there has been in the ~15 years lifespan of this game, it will always come down to what statistic you highlight and what metric you choose. But sadly, your way just wasn't solid at all. It was highly biased and too radical to correctly adjust for change (not that you particularly need to). But the funniest part is honestly that you think your approach is completly unbiased... Yeah, this measurement is quite faulty I think, but it brings a great point across which ironically is the one OP quoted in his post "If you torture the data long enough, it will confess to anything" At the end of the day you can use all the fancy stats and data you want, but it doesn't make your ranking any less subjective than someone's guts feeling, whether it's premobeats way overvaluing things like winrates or aligulac rating to prob Serral or the thing ejozl did here. And we all know the one true Goat ranking is this one: 1. Serral 2. Maru (extremely close to Serral) 3. Inno 4. Rogue 5. Life 6. Zest 7. sOs 8. herO 9. Dark 10. Stats I think that quote was put there unironically to highlight the sarcastic over-the-top nature of the article. But why do you mischaracterize what was happening? I wrote in the article that I did not come up with this weighting. I even posted the exact quote that I inserted into ChatGTP so that everyone can check it. Since then, I admitted several times that the weighting is off and even proposed a different one, where Aligulac and match win rates only make up 5% each. Also I never propped Serral lol. He gets penalized the most by subjective decisions I had to make (including non-prime years in tournament win percentage and leave out the inflation-bonus in the match win rate). I remember being asked in the comments of my first article, if I would post the results, if changing the methodology of the original article would mean that Serral would not come out on top in any of the metrics. Well, as it is obvious now, I have no issue accepting that Serral is not the best in tournament win percentage. But it seems like people are trying to critique my result without actually addressing any core methodological feature (except for weighting, which was explained weeks ago). What I would find more troubling than obvious weighting issues is a GOAT ranking that presents itself as statistical but in the end is a post-hoc feeling based scripture. And of course you can make a ranking that is a lot more objective than someone's gut feeling. Not being in alignment with every subjective feeling for era-difference or tournament distinctions does not mean that going as close as possible in a statistical evaluation is the same as a random StarCraft II player that thinks this or that player is the GOAT. We can very clearly see from the statistical evaluation that Serral is way ahead of Maru and Rogue in the time-frame that they played together. We can see that Life was ahead of Rain. We can also see what everyone subjectively thought for a long time: That Rogue has pretty mixed results. Plus, subjectivity can be put into a bare minimum. 1. We need a point/ranking system for the tournaments (similar to what Miz and I did) 2. We need some kind of era-multiplier for different time brackets in terms of how hard it was to win (This can also be calculated more or less ok-ish as well) OR come to the conclusion that all time periods were the same, which I doubt is realistic or correct 3. Decide which metrics are the most important and then add a weighting (the community here and on reddit already agree that tournament score and tournament win participation are the most important These are all the subjective things one needs and the community will probably come to a close proximity. But, even in the case that we have heavy disagreements here: When I revise my method one more time, I will try to make it in a way that people can tell me their desired ratios and I can post the result based on these suggestions. When all my excels are interlinked, I can easily take community suggestions for these three and we can look at how hard we would need to torture the data to make this or that player come out on top. You do not penalize Serral, though I think we all understand rhetorically why you say you do. You give him bonuses. A lot of them. And you only admit it when people call out the more obvious and ridiculous ways in which you do it like your 20% Aligulac weighting. The basic logical error you make is that you are calling all your criteria modifiers "penalties" to Serral but whether they are penalties or bonuses entirely depends on baselines versus expectations in your modifiers. Take your tournament win % criteria. Again you call your 20% (or whatever it was) modifier a "penalty" to Serral. But the question is whether that is too high or too low or just right given that Serral competed exclusively in tournaments that are significantly easier to win than KILs, which have dozens and sometimes hundreds of additional competitors, harder groups stages, more group stages, etc. If it's just right, there's no penalty or bonus to Serral, if it's too high, it's a penalty to Serral, and if it's too low, it's a bonus to Serral. I simply do not agree that the 20% discount reflects the comparative odds of winning a 2020 Dreamhack Masters versus a 2013 GSL or SSL. It's probably 5X easier to win the former than the latter imho. In this specific example, though, it doesn't even matter because it's actually a bonus to Serral irrespective of whether the 20% modifier is right or wrong as he is the only player on your GOAT list that does not have his tournament win % diluted by participation in the signiifcantly-harder-to-win KILs. This is only one example of how you claim to be penalizing Serral when you're actually giving him bonuses in your model. This kind of logical error permeates your model. At first I was giving you the benefit of the doubt that you did not understand some of these concepts in predictive modeling, but it's pretty clear that you do and are just hoping that others do not understand so you can get away with saying over and over again that you are penalizing Serral. Your entire goal is to say that no matter how much you penalize Serral, he is the GOAT, and because you're so focused on this, it's producing many errors and requiring you to go to great lengths to obscure what it is you are doing. The simple fact is that your model giving Serral 2X the GOAT points of the next highest GOAT and 5X the GOAT points of Rogue is on its face not credible. You try to get around this by hiding behind your methodology, but in just a very cursory review of your methodology like I've done here reveals huge errors, almost all of which are designed to give Serral bonus GOAT points. Wouldn't it be better at this point just to admin you've created a Serral = GOAT calculator? It's totally legitimate to call Serral the GOAT--he's probably most people's GOAT, even most players probably think he's the GOAT. You created a highly subjective model to crown Serral the GOAT. It's no different than this GOAT list, just different subjective criteria. It is not rhetorical to say that I penalized Serral, because I did: 1a. Do you deny that the Koreans played more lower tier players in comparison to Serral? Or do you deny that I wrote that I did not include this correction, which is detrimental to Serral? 1b. Do you deny that including non-prime years, which I did, penalizes/hurts Serral the most according to the data? 2. In the article - as well as multiple times in the other thread (and even in the very post you commented on and directly to you in the other thread!!) I explained that this weighting was proposed by ChatGTP. I still don't understand why you keep pounding on that, tbh. As WombaT wrote out in the other thread, the weighting is entirely superfluous. The most important revelation was that most players have mixed results, while Serral is dominating them all. You make it seem like the weighting was there to push Serral, which is not the case. You have no proof for this accusation. On the contrary, I posted the exact text that I used to let ChatGTP decide on the weighting, to let you guys double check. I only did so because I knew based on the results, that the weighting does not change in the first place and I couldn’t decide on a specific weighting, which I explained ad nauseam in the article as well as in the comments. First place would only change with an insane weighting-setup for efficiency… then Life (who most people disregard anyway) would be on top, with Serral being 2nd. Arguing with the weighting result is utterly pointless… As this is your (and other critics) biggest point which you all come back to every time, I doubt that is much more substance to gain from you guys. 3a. In the tournament win %, similar to most metrics, I discounted all region-locked tournaments. This is a penalty to Serral, although a needed one to deny his easier tournament wins. But still it is a penalty; especially to deny them entirely, as Korean pros had rather easy tournament wins as well. 3b. You might disagree that the bonus is not enough according to your subjective feeling. My reasoning for using 20%: I explained in detail why a larger player pool, especially in mostly tier 2, 3 and 4 players, would not decrease a tournament win percentage that much. - doubling the player pool would most likely mean just one extra round. If that round was a group stage, it would naturally mean that you’d have to play more lower tier player statistically, boosting your match win rate - I reasoned that this fact means that average place would need a higher era-multiplier than tournament win percentage - Now looking at Serral’s average win rate of around 82% in his years from 2018-2024 subtracted with 2% (some lower match win for era but also considering that his win rates are deflated versus the Koreans because of 1a, also taking into account that earlier group stages are easier to win than knockout brackets), these more rounds would mean a win probability of roughly 80%. As an example, going from 4 rounds to 5 rounds would statistically mean a win probability 40% to 33%, meaning a loss of 17,5% or a boost of 21,2%. - Now, comparing DH 2020 to a GSL or SSL 2013: Not all tournaments that other players have won were a GSL though. What about the 2013 Hot6ix Cup or ASUS ROG NorthCon? And didn’t Serral win uberly stacked tournaments as well? You are committing a false equivalence/improper comparison fallacy here. In essence, you are right that one can’t compare a GSL to a DH. But that is also not what I did. There were not only GSLs in 2013 (as a matter of fact, there was only one) and Serral won more stacked tournaments as well. It also borders on a strawman, as my approach was not designed to precisely reflect the odds between only one of the hardest 2013 tournaments vs a specific 2020 event, but rather roughly corrects across periods. - Do you further realize that arguing like this from an era/period-perspective also argues against Rogue’s and Maru’s case? - What adjustment would you find appropriate? Overall it is rather the modern Koreans that benefitted in this metric, as Serral, Clem and mostly Reynor were missing in the GSLs that they won. Of course Serral wouldn’t have the same tournament win percentage had he been a full-time pro at his peak playing in the prime era. But the win percentages of INno, Rain, etc. would drop as well, had Mvp, Serral and Clem played there. I explained this already in the article. “Again you call your 20% (or whatever it was) modifier a "penalty" to Serral” Can you give the exact quote? This would be much more fruitful if you didn’t try to rely on notions like my method was not statistical, keep using arguments that have been addressed or straight out deny that Serral was penalized on several occasions, which he clearly was. Talk about the points ratio in the tournaments or the era-multiplier or your preferred weighting. Or - as you by your own account don’t care for statistical GOAT debates anyway - simply stay out of them. But making unsubstantiated accusations like me making a Serral GOAT-calculator or repeating the same already addressed notions is absolutely ridiculous. You "penalize" Serral within metrics that heavily favor him. For comparison, it's like if I would incorporate a metric for number of seperate years with a playoffs finish (a metric obviously favoring Maru since he had the longest career), giving it 25% of the overall score, and then acting like I penalized Maru because I considered korean tournaments the same as foreigner-only tournaments, despite being obviously harder. I already said in the article that there are a million ways to fragment the data... one could also make a scoreboard for best cannon rush defender or what you just suggested. But it simply makes no sense, as I used data that covers the entirety of the game and metrics that are most valuable to determine the Greatest of All Time. If you disagree, simply explain why your example is a worthy metric. But your comparison is completely off, as the tournament win percentage is a simple data set. Nothing of it has been modified like you did. The only thing that was modified was - disregarding region locks, which is a (needed) penalty for Serral - add a multiplier that boosted everything pre-2018 and penalized everything post-2018 (which penalized Serral among others) So which metric does favor Serral? And why exactly does it "favor" him? If it's entirely reasonable to modify the data to enable comparability, I wouldn't call it a "penalty". If we're talking about tournament wins for example, the raw data set should include online cups, however as we both agree it makes no sense to include them in a GOAT analysis and exclude them. Despite that I wouldn't call it a Clem "penalty", it just removes data which is not relevant for the topic. That's how I feel about your so called Serral "penalty".
So if I remove these tournaments for Serral, shouldn't I check as well for tournaments where Koreans had it similarly easy? Because I did not do that... hence, it was a penalty. Although I get your overall point here, which is true: A contextual adjustment is not a penalty.
EDIT: I agree that some metrics show similar things... hence I proposed a 5% split for the not so relevant ones. We can also go 3% or I can replicate Miz' 0,55 to 0,45 (Tournament score versus tournament win percentage). As I said.. I want to make it completely transparanet for the next time around, in terms of checking for different weightings.
|
I mean you are both arguing that Serral is the GOAT with the only difference by how large a margin.
|
On July 15 2025 22:06 Harris1st wrote:
I mean you are both arguing that Serral is the GOAT with the only difference by how large a margin.
I am technically arguing for the fact that I didn't favor anyone in my analysis and honestly don't care if this or that player comes out on top...
From the last article to the newer one I took in community suggestions like including Team League and including more players. I even made further data based analyses after Charoisaur's comments in regards to the Korean-deterioration-argument.
It simply is irritating to me that I get accused of favoritism or being intellectually incapable of understanding a point, when no sound arguments are presented.
So far, the only thing rwala suggested that wasn't borderline/directly insulting or superfluous, was the fact the era-multiplier for the tournament win percentage is off. I mean... we could obviously discuss, if 20% is too much, too little or approximately fine but the whole scenery of trying to paint me as a die-hard Serral-fanboy surrounding this pretty easy notion is utterly redundant.
I further think Charoisaur is correct in pointing out that some of my metrics take care of rather similar feats... but that is exactly why I already proposed a different weighting that focuses on the tournament score and tournament win percentage as the two most important metrics and even suggested that the other 4 metrics could be minimized even more.
|
On July 16 2025 01:27 PremoBeats wrote:Show nested quote +On July 15 2025 22:06 Harris1st wrote:
I mean you are both arguing that Serral is the GOAT with the only difference by how large a margin.
trying to paint me as a die-hard Serral-fanboy
Well, you certainly are lol, we can all see your post history.
That doesn't mean that every ranking you ever come up with should be instantly disregarded but in your first iteration the Serral bias was definitely far too strong and obvious.
|
On July 16 2025 01:59 Charoisaur wrote:Show nested quote +On July 16 2025 01:27 PremoBeats wrote:On July 15 2025 22:06 Harris1st wrote:
I mean you are both arguing that Serral is the GOAT with the only difference by how large a margin.
trying to paint me as a die-hard Serral-fanboy Well, you certainly are lol, we can all see your post history. That doesn't mean that every ranking you ever come up with should be instantly disregarded but in your first iteration the Serral bias was definitely far too strong and obvious.
I was using strong language because I was hyped by the results... I even acknowledged that the hype-language was over the top. Perhaps I became a slight fanboy then, who knows. But if it happened, it was because of the results, that still are pretty robust. To quote WombaT from another thread: "I try to be as unbiased as I can, I think I don’t appear so merely because others are so biased that an attempt at neutrality appears biased to them."
I ask you again: Do you have any criticism about methodology except the ones we already discussed? Did I understand your metric proposal correctly that it is basically pretty much the same as my tournament score looked at through a yearly lense? A method which I didn't apply for any other metric that went into the final result?
|
|
|
|
|
|
EPT
