|
On June 01 2025 23:54 Hider wrote:Show nested quote +On June 01 2025 16:16 Mizenhauer wrote:
My only question is, given that Rogue's GOAT score is so disproportionately low compared to the other players in this study, do you feel it's an indication of some fundamental error in your process? I say this because that kind of skew in results leads me to believe that something is inherently incorrect when it comes to the calculation/tabulation of your GOAT score. Maybe that's ChatGPT (it's pretty damning). Isn't the thing about Rogue that he was more streaky or perhaps more clutchy. I never looked at Rogue and thought he was dominant or even the best player in the world at any point in time. Personally I put MVP/Inno over Rogue on my GOAT list. A criteria for being the GOAT is imo that you were the best player in the world for some period in time. I weight MVP's era of dominance and the early KESPA era multiple times higher per tournament than what happened in the past few years. On the other hand we had more tournaments back then so that offsets it partially. On a per tournament basis I would give perhaps 5 times more weight to a tournament during Starcraft's peak era than today. However, the margin of dominance also matters, and that's where Serral truly shines and still comes in at number 1. MVP had perhaps 6 months where he looked a level above everyone else, but after that it felt more like he was one of the better players but just pretty clutchy. In contrast Serral has always been the favorite or huge favorite for 7 years straight.
5 times more? Like a blank multiplier? Even though there were Premier Tournaments where two to three top tier Koreans flew to foreign regions to farm them? These situations are comparable to Serral winning in Europe and you'd want to award these tournaments 5 times the amount in relation to modern ones?
I don't think that this is a sound multiplier, unless I misunderstood you.
On June 02 2025 00:30 JJH777 wrote:
Aligulac being 20% of your determination is insane. Aligulac is a bad ELO system for a bunch of reasons. It consistently overrates people from outside of the KR scene. There are tons of examples even before Serral was notable. Serral specifically hit rank 1 for the first time when he still had <50% win rate vs Koreans. For much of his career his ranking was inflated by playing a huge amount of games vs mid to low tier EU players which he always swept. Maru would have swept those too. If you put Maru into 3 8 player round robin groups vs non Serral/Reynor/Clem EU players per year (format of EU regional for most of the years being discussed) what do you think happens to his aligulac rating? I guarantee it skyrockets. It's also funny that you punish Marus tournament winrrate for winning against "low tier" Koreans but don't take that into account that Serral's aligulac ranking heavily benefits from beating up low tier EU players.
I also heavily disagree with the idea of average placement/efficiency/percentage of won tournaments in all premiers played being important. That just specifically punishes players with long careers who consistently qualified for premiers. A longer career with consistent premier qualifications should be a boon to goat candidacy not a negative. The fact that Marus ranking in these categories would go up dramatically if you exclude his pre 2018 results shows alone how flawed it is. Maru being an S tier player from 2013-2018 should not penalize his chances of being goat.
I know for some of these aspects you slightly adjusted for this by determining prime years but I disagree with it being a consideration at all. Certainly not a combined 42.5% of your ranking.
1. The overall result wouldn't change (Serral would still lead with nearly 300 points), even if I removed the Aligulac-metric entirely
2. I think you got this the wrong way round. The machine's algorithm saw the explosion of wins Serral had versus Europeans and through cross-regional comparisons correctly "predicted" that he would defeat the Koreans as well, which was exactly what happened in the first quarter of 2018 when his win rates skyrocketed to over 80%. What you mentioned is a super rare case, that shows how well Aligulac works cross-regionally. These inflation-arguments were a thing in the early 2010s... not anymore in 2017.
Further, after 2018, Serral played enough Koreans per year, which made the calculation pretty reliable.
3. I only pointed the Korean's inflation against Serral through weaker Korean players out. As I wrote in the article, I did not actually incorporate these findings in the calculation.
4. No, it punishes no one. It simply shows, that there are some players that have the ability to always finish on top and others that don't. That the player you favor by chance is not good in this metric, does not mean that the metric per se is bad.
5. Some argue rightfully, that it would be unfair to players like Rain and Mvp who had short, kick-started careers if one discounts the worse years of players, hence I went for a mix. In the tournament score Maru's longer careers ears him more points. And if he was as efficient as Serral over 7 years, he'd easily be up there with him. That is simply what a fair analysis shows... strengths and weaknesses.
|
On June 01 2025 23:54 Hider wrote:Show nested quote +On June 01 2025 16:16 Mizenhauer wrote:
My only question is, given that Rogue's GOAT score is so disproportionately low compared to the other players in this study, do you feel it's an indication of some fundamental error in your process? I say this because that kind of skew in results leads me to believe that something is inherently incorrect when it comes to the calculation/tabulation of your GOAT score. Maybe that's ChatGPT (it's pretty damning). Isn't the thing about Rogue that he was more streaky or perhaps more clutchy. I never looked at Rogue and thought he was dominant or even the best player in the world at any point in time. Personally I put MVP/Inno over Rogue on my GOAT list. A criteria for being the GOAT is imo that you were the best player in the world for some period in time. I weight MVP's era of dominance and the early KESPA era multiple times higher per tournament than what happened in the past few years. On the other hand we had more tournaments back then so that offsets it partially. On a per tournament basis I would give perhaps 5 times more weight to a tournament during Starcraft's peak era than today. However, the margin of dominance also matters, and that's where Serral truly shines and still comes in at number 1. MVP had perhaps 6 months where he looked a level above everyone else, but after that it felt more like he was one of the better players but just pretty clutchy. In contrast Serral has always been the favorite or huge favorite for 7 years straight.
Rogue absolutely was the best player in the world in late 2017 - early 2018. There's absolutely no argument about it, he won IEM Shanghai, GSL Super tournament and Blizzcon back-to-back-to-back, then got eliminated from GSL but won IEM Katowice right after. It's one of the most dominant streaks any player ever had
|
All those words and analysis and then it leads into this.
These are the weightings CHAT GTP suggested:
Aligulac Rank: 20%
Match Win Rate: 15%
Tournament win %: 17,5%
Average placement: 15%
Tournament Score: 22,5%
Efficiency Score: 10%
Although I don’t fully agree with all of the suggested weights, I adopted them as-is - both for simplicity and because the final outcome would only shift significantly under an extreme weighting of efficiency.
If you want to be taken serious, you should have a better reasoning for your weightings than because chatgpt said so.
Personally I don't see any reason why any score besides tournament score should count AT ALL. tournament score is what players have achieved during their career which is what the Goat is about imo. The rest are just extremely subjective determinators of things like 'dominance'.
But I don't see any good reason why dominance should count more then longevity aka why should it count more when a player wins 8 tournaments during two years (and thus has good scores in your chosen criteria) versus when he needs 15 years to win 8 tournaments. The latter shows extreme longevity staying competitive for so long, and succeeding in vastly different metas, environments and competition, which to me is just as impressive as being dominant during one stint but failing to keep that performance for a longer period.
But 5 out of your 6 criteria heavily favor dominance in a shortish period over longevity which I just can't agree with.
Rating the actual tournament results of a player with only 22.5% seems completely asinine to me
|
On June 02 2025 01:06 PremoBeats wrote:Show nested quote +On June 01 2025 23:54 Hider wrote:On June 01 2025 16:16 Mizenhauer wrote:
My only question is, given that Rogue's GOAT score is so disproportionately low compared to the other players in this study, do you feel it's an indication of some fundamental error in your process? I say this because that kind of skew in results leads me to believe that something is inherently incorrect when it comes to the calculation/tabulation of your GOAT score. Maybe that's ChatGPT (it's pretty damning). Isn't the thing about Rogue that he was more streaky or perhaps more clutchy. I never looked at Rogue and thought he was dominant or even the best player in the world at any point in time. Personally I put MVP/Inno over Rogue on my GOAT list. A criteria for being the GOAT is imo that you were the best player in the world for some period in time. I weight MVP's era of dominance and the early KESPA era multiple times higher per tournament than what happened in the past few years. On the other hand we had more tournaments back then so that offsets it partially. On a per tournament basis I would give perhaps 5 times more weight to a tournament during Starcraft's peak era than today. However, the margin of dominance also matters, and that's where Serral truly shines and still comes in at number 1. MVP had perhaps 6 months where he looked a level above everyone else, but after that it felt more like he was one of the better players but just pretty clutchy. In contrast Serral has always been the favorite or huge favorite for 7 years straight. 5 times more? Like a blank multiplier? Even though there were Premier Tournaments where two to three top tier Koreans flew to foreign regions to farm them? These situations are comparable to Serral winning in Europe and you'd want to award these tournaments 5 times the amount in relation to modern ones? I don't think that this is a sound multiplier, unless I misunderstood you. Show nested quote +On June 02 2025 00:30 JJH777 wrote:
Aligulac being 20% of your determination is insane. Aligulac is a bad ELO system for a bunch of reasons. It consistently overrates people from outside of the KR scene. There are tons of examples even before Serral was notable. Serral specifically hit rank 1 for the first time when he still had <50% win rate vs Koreans. For much of his career his ranking was inflated by playing a huge amount of games vs mid to low tier EU players which he always swept. Maru would have swept those too. If you put Maru into 3 8 player round robin groups vs non Serral/Reynor/Clem EU players per year (format of EU regional for most of the years being discussed) what do you think happens to his aligulac rating? I guarantee it skyrockets. It's also funny that you punish Marus tournament winrrate for winning against "low tier" Koreans but don't take that into account that Serral's aligulac ranking heavily benefits from beating up low tier EU players.
I also heavily disagree with the idea of average placement/efficiency/percentage of won tournaments in all premiers played being important. That just specifically punishes players with long careers who consistently qualified for premiers. A longer career with consistent premier qualifications should be a boon to goat candidacy not a negative. The fact that Marus ranking in these categories would go up dramatically if you exclude his pre 2018 results shows alone how flawed it is. Maru being an S tier player from 2013-2018 should not penalize his chances of being goat.
I know for some of these aspects you slightly adjusted for this by determining prime years but I disagree with it being a consideration at all. Certainly not a combined 42.5% of your ranking. 1. The overall result wouldn't change (Serral would still lead with nearly 300 points), even if I removed the Aligulac-metric entirely 2. I think you got this the wrong way round. The machine's algorithm saw the explosion of wins Serral had versus Europeans and through cross-regional comparisons correctly "predicted" that he would defeat the Koreans as well, which was exactly what happened in the first quarter of 2018 when his win rates skyrocketed to over 80%. What you mentioned is a super rare case, that shows how well Aligulac works cross-regionally. These inflation-arguments were a thing in the early 2010s... not anymore in 2017. Further, after 2018, Serral played enough Koreans per year, which made the calculation pretty reliable. 3. I only pointed the Korean's inflation against Serral through weaker Korean players out. As I wrote in the article, I did not actually incorporate these findings in the calculation. 4. No, it punishes no one. It simply shows, that there are some players that have the ability to always finish on top and others that don't. That the player you favor by chance is not good in this metric, does not mean that the metric per se is bad. 5. Some argue rightfully, that it would be unfair to players like Rain and Mvp who had short, kick-started careers if one discounts the worse years of players, hence I went for a mix. In the tournament score Maru's longer careers ears him more points. And if he was as efficient as Serral over 7 years, he'd easily be up there with him. That is simply what a fair analysis shows... strengths and weaknesses.
Aligulac called King Kong a top 10 player in the world as late as 2014. I'm pretty sure he has never beaten a top pro. He got that ranking by being dominant in his region while never having any global results. He stayed in the top 50 range for years despite the fact that on a skill basis he was likely never even top 100. Aligulac has huge issues judging players who dominate their own region. Another good example is Polt. Polt is in the top 15 all time on aligulac. He was regularly top 5 and even 1st while active. He did that by dominating NA. He was never truly competitive with top GSL players post 2011 but dominating a weaker region kept him at the top of aligulac for ages.
Saying aligulac was just predicting Serral's future success is a copout when that future success wasn't until over a year after he hit rank 1. During which time he regularly lost matches to Koreans but still during that entire period he was almost always 1st with just a few falls down to 2-3. I'm definitely not saying he didn't deserve it the majority of the time he had it but the margin he had it by was heavily inflated and there were many times he would have been passed with a system that didn't over reward winning against far weaker competition.
I don't know how possible it is to calculate, probably depends on how accessible aligulac's source code/algorithm and result sets are but I am genuinely curious about how aligulac would have changed if Maru played 24 bo3s vs mid and lower rank EU players (basically everyone sans Serral and Reynor until 2022 then everyone besides them and Clem after that) every year from 2017-2024. Giving him whatever his current winrrate is vs that group in those bo3s. Based on aligulac's history of over rewarding beating lower competition I am confident his rating would increase significantly but I can't say for sure. I may try to do that someday when I have free time.
Marus "ability to finish on top" is punished by playing in an era with far more active full time pros and by having a far higher amount of events played in. Those criteria definitely uniquely punishes Marus career as it is the longest of any top pro by a considerable margin. By those criteria he would literally rank higher if he had just failed to qualify for events than if he qualifies for them. The fact that you can't agree that is backwards is crazy to me.
Serral's metrics in all categories besides tournament results are inflated by not playing in GSL. If Serral played in GSL regularly it's extremely likely his aligulac, matchup win rates, efficiency, average tournament placement, and percentage of won tournaments would all decline. I do agree he would likely win some GSLs which would benefit his tournament score but when you're only giving tournament score 22% of the overall weight it's not like that makes much difference.
|
5 times more? Like a blank multiplier? Even though there were Premier Tournaments where two to three top tier Koreans flew to foreign regions to farm them? These situations are comparable to Serral winning in Europe and you'd want to award these tournaments 5 times the amount in relation to modern ones?
I don't think that this is a sound multiplier, unless I misunderstood you.
Tournaments should all be adjusted for difficulty. Aligulac ratings could be used for that. If a tournament is relatively easy to win for Serral as most top 10-15 players aren't competiting it should reward fewer points. Hence Serral would get less points for that - all things equal.
But noone aside from MVP would really benefit that much as I also rate "dominance" much higher than most other people do. And noone was really clearly dominant for a very long time during those periods.
However, assuming say someone was equally dominant from 2011 to 2013 as Serral has been past 7 years, he would rank above Serral in my book.
I don't know exactly which tournaments in korea 2011-2014 that would get a lot of points assuming they are adjusted for skill which you don't think should receive that many points?
|
On June 02 2025 02:18 JJH777 wrote:Show nested quote +On June 02 2025 01:06 PremoBeats wrote:On June 01 2025 23:54 Hider wrote:On June 01 2025 16:16 Mizenhauer wrote:
My only question is, given that Rogue's GOAT score is so disproportionately low compared to the other players in this study, do you feel it's an indication of some fundamental error in your process? I say this because that kind of skew in results leads me to believe that something is inherently incorrect when it comes to the calculation/tabulation of your GOAT score. Maybe that's ChatGPT (it's pretty damning). Isn't the thing about Rogue that he was more streaky or perhaps more clutchy. I never looked at Rogue and thought he was dominant or even the best player in the world at any point in time. Personally I put MVP/Inno over Rogue on my GOAT list. A criteria for being the GOAT is imo that you were the best player in the world for some period in time. I weight MVP's era of dominance and the early KESPA era multiple times higher per tournament than what happened in the past few years. On the other hand we had more tournaments back then so that offsets it partially. On a per tournament basis I would give perhaps 5 times more weight to a tournament during Starcraft's peak era than today. However, the margin of dominance also matters, and that's where Serral truly shines and still comes in at number 1. MVP had perhaps 6 months where he looked a level above everyone else, but after that it felt more like he was one of the better players but just pretty clutchy. In contrast Serral has always been the favorite or huge favorite for 7 years straight. 5 times more? Like a blank multiplier? Even though there were Premier Tournaments where two to three top tier Koreans flew to foreign regions to farm them? These situations are comparable to Serral winning in Europe and you'd want to award these tournaments 5 times the amount in relation to modern ones? I don't think that this is a sound multiplier, unless I misunderstood you. On June 02 2025 00:30 JJH777 wrote:
Aligulac being 20% of your determination is insane. Aligulac is a bad ELO system for a bunch of reasons. It consistently overrates people from outside of the KR scene. There are tons of examples even before Serral was notable. Serral specifically hit rank 1 for the first time when he still had <50% win rate vs Koreans. For much of his career his ranking was inflated by playing a huge amount of games vs mid to low tier EU players which he always swept. Maru would have swept those too. If you put Maru into 3 8 player round robin groups vs non Serral/Reynor/Clem EU players per year (format of EU regional for most of the years being discussed) what do you think happens to his aligulac rating? I guarantee it skyrockets. It's also funny that you punish Marus tournament winrrate for winning against "low tier" Koreans but don't take that into account that Serral's aligulac ranking heavily benefits from beating up low tier EU players.
I also heavily disagree with the idea of average placement/efficiency/percentage of won tournaments in all premiers played being important. That just specifically punishes players with long careers who consistently qualified for premiers. A longer career with consistent premier qualifications should be a boon to goat candidacy not a negative. The fact that Marus ranking in these categories would go up dramatically if you exclude his pre 2018 results shows alone how flawed it is. Maru being an S tier player from 2013-2018 should not penalize his chances of being goat.
I know for some of these aspects you slightly adjusted for this by determining prime years but I disagree with it being a consideration at all. Certainly not a combined 42.5% of your ranking. 1. The overall result wouldn't change (Serral would still lead with nearly 300 points), even if I removed the Aligulac-metric entirely 2. I think you got this the wrong way round. The machine's algorithm saw the explosion of wins Serral had versus Europeans and through cross-regional comparisons correctly "predicted" that he would defeat the Koreans as well, which was exactly what happened in the first quarter of 2018 when his win rates skyrocketed to over 80%. What you mentioned is a super rare case, that shows how well Aligulac works cross-regionally. These inflation-arguments were a thing in the early 2010s... not anymore in 2017. Further, after 2018, Serral played enough Koreans per year, which made the calculation pretty reliable. 3. I only pointed the Korean's inflation against Serral through weaker Korean players out. As I wrote in the article, I did not actually incorporate these findings in the calculation. 4. No, it punishes no one. It simply shows, that there are some players that have the ability to always finish on top and others that don't. That the player you favor by chance is not good in this metric, does not mean that the metric per se is bad. 5. Some argue rightfully, that it would be unfair to players like Rain and Mvp who had short, kick-started careers if one discounts the worse years of players, hence I went for a mix. In the tournament score Maru's longer careers ears him more points. And if he was as efficient as Serral over 7 years, he'd easily be up there with him. That is simply what a fair analysis shows... strengths and weaknesses. Aligulac called King Kong a top 10 player in the world as late as 2014. I'm pretty sure he has never beaten a top pro. He got that ranking by being dominant in his region while never having any global results. He stayed in the top 50 range for years despite the fact that on a skill basis he was likely never even top 100. Aligulac has huge issues judging players who dominate their own region. Another good example is Polt. Polt is in the top 15 all time on aligulac. He was regularly top 5 and even 1st while active. He did that by dominating NA. He was never truly competitive with top GSL players post 2011 but dominating a weaker region kept him at the top of aligulac for ages. Saying aligulac was just predicting Serral's future success is a copout when that future success wasn't until over a year after he hit rank 1. During which time he regularly lost matches to Koreans but still during that entire period he was almost always 1st with just a few falls down to 2-3. I'm definitely not saying he didn't deserve it the majority of the time he had it but the margin he had it by was heavily inflated and there were many times he would have been passed with a system that didn't over reward winning against far weaker competition. I don't know how possible it is to calculate, probably depends on how accessible aligulac's source code/algorithm and result sets are but I am genuinely curious about how aligulac would have changed if Maru played 24 bo3s vs mid and lower rank EU players (basically everyone sans Serral and Reynor until 2022 then everyone besides them and Clem after that) every year from 2017-2024. Giving him whatever his current winrrate is vs that group in those bo3s. Based on aligulac's history of over rewarding beating lower competition I am confident his rating would increase significantly but I can't say for sure. I may try to do that someday when I have free time. Marus "ability to finish on top" is punished by playing in an era with far more active full time pros and by having a far higher amount of events played in. Those criteria definitely uniquely punishes Marus career as it is the longest of any top pro by a considerable margin. By those criteria he would literally rank higher if he had just failed to qualify for events than if he qualifies for them. The fact that you can't agree that is backwards is crazy to me. Serral's metrics in all categories besides tournament results are inflated by not playing in GSL. If Serral played in GSL regularly it's extremely likely his aligulac, matchup win rates, efficiency, average tournament placement, and percentage of won tournaments would all decline. I do agree he would likely win some GSLs which would benefit his tournament score but when you're only giving tournament score 22% of the overall weight it's not like that makes much difference.
- I already said that Aligulac had issues in the 2010s.
- Didn't he hit rank 1 on the last list of 2017? And started winning against Koreans in January 2018? I mean.. that is exactly what the machine is built to do. Weigh players relative to their statistics against each other.
- Which exact time frame are you talking about when talking about Maru? And which metric?
- Why would Serral lose more games in GSL than versus top Koreans at other Premier Tournaments? Because of its structure?
On June 02 2025 02:15 Charoisaur wrote:All those words and analysis and then it leads into this. Show nested quote +These are the weightings CHAT GTP suggested:
Aligulac Rank: 20%
Match Win Rate: 15%
Tournament win %: 17,5%
Average placement: 15%
Tournament Score: 22,5%
Efficiency Score: 10%
Although I don’t fully agree with all of the suggested weights, I adopted them as-is - both for simplicity and because the final outcome would only shift significantly under an extreme weighting of efficiency. If you want to be taken serious, you should have a better reasoning for your weightings than because chatgpt said so. Personally I don't see any reason why any score besides tournament score should count AT ALL. tournament score is what players have achieved during their career which is what the Goat is about imo. The rest are just extremely subjective determinators of things like 'dominance'. But I don't see any good reason why dominance should count more then longevity aka why should it count more when a player wins 8 tournaments during two years (and thus has good scores in your chosen criteria) versus when he needs 15 years to win 8 tournaments. The latter shows extreme longevity staying competitive for so long, and succeeding in vastly different metas, environments and competition, which to me is just as impressive as being dominant during one stint but failing to keep that performance for a longer period. But 5 out of your 6 criteria heavily favor dominance in a shortish period over longevity which I just can't agree with. Rating the actual tournament results of a player with only 22.5% seems completely asinine to me
1 criteria is for efficiency, as the name suggests. Aligulac is for dominance and longevity. Tournament win % does not exclude longevity, as Serral shows. Average placement also does not exclude it. Neither do match win rates. Tournament score is also both.
Most of the metrics can show both, although some are harder to sustain for long periods, I agree.
But you do realize that the weighting is mostly irrelevant, unless we hyper-overtune efficiency, so Life would come out on top? Serral simply is good in all of them. For all I care give each of them 0,05 and the rest to the tournament score. Then Maru and Rogue would be at the same place as in Miz' list.
@Hider: That is exactly what I did though. I looked at each (!!!) tournament and looked at the Ro8 and Ro16, noted down every player and checked the corresponding list in order to be able to build an average. According to this average and more subjective factors like prestige and prize money I established 7 categories.
|
Most of the metrics can show both, although some are harder to sustain for long periods, I agree.
Yeah, that's the point. It just heavily punishes players who have high longevity but were never really dominant (like Dark), compared to players who only shined during a brief period but were very dominant (like Mvp).
I don't see any objective reasons why the achievements of someone like Dark should be worth less, just because he needed more time to accomplish them. In the opposite, he proved he can succeed in many more different metas and fields of competition than Mvp, which should count for something.
It's a good ranking for most dominant player of all time, but at least for me that's not (solely) what Greatest of all time means.
|
On June 02 2025 04:26 Charoisaur wrote:Show nested quote +Most of the metrics can show both, although some are harder to sustain for long periods, I agree. Yeah, that's the point. It just heavily punishes players who have high longevity but were never really dominant (like Dark), compared to players who only shined during a brief period but were very dominant (like Mvp). I don't see any objective reasons why the achievements of someone like Dark should be worth less, just because he needed more time to accomplish them. In the opposite, he proved he can succeed in many more different metas and fields of competition than Mvp, which should count for something. It's a good ranking for most dominant player of all time, but at least for me that's not (solely) what Greatest of all time means.
So what's the suggestion? Only look at prime years for a comparison? Wouldn't that be unfair to players like Rain and Mvp who kickstarted into the scene?
Also, why should it be done in the first place, when we have a player who demonstrates that it is possible to do. Could be interpretated as favoritism towards Inno and Maru. Do we subsequently also discount those years in the tournament score though?
I could do it for a comparison though... this wouldn’t take too much time I guess, as I'd simply have to redo the prime years, which shouldn't take too long.
Edit: I already did for average place anyways, as those data sets were super bothersome to gather.
|
On June 02 2025 04:40 PremoBeats wrote:Show nested quote +On June 02 2025 04:26 Charoisaur wrote:Most of the metrics can show both, although some are harder to sustain for long periods, I agree. Yeah, that's the point. It just heavily punishes players who have high longevity but were never really dominant (like Dark), compared to players who only shined during a brief period but were very dominant (like Mvp). I don't see any objective reasons why the achievements of someone like Dark should be worth less, just because he needed more time to accomplish them. In the opposite, he proved he can succeed in many more different metas and fields of competition than Mvp, which should count for something. It's a good ranking for most dominant player of all time, but at least for me that's not (solely) what Greatest of all time means. So what's the suggestion? Only look at prime years for a comparison? Wouldn't that be unfair to players like Rain and Mvp who kickstarted into the scene? Also, why should it be done in the first place, when we have a player who demonstrates that it is possible to do. Could be interpretated as favoritism towards Inno and Maru. Do we subsequently also discount those years in the tournament score though? I could do it for a comparison though... this wouldn’t take too much time I guess, as I'd simply have to redo the prime years, which shouldn't take too long. Edit: I already did for average place anyways, as those data sets were super bothersome to gather.
I already said it, for me the tournament score is the only metric that's relevant because those are the actual achievements a player has accomplished during their career. I don't see why this score needs to be adjusted with other, very subjective metrics. Accomplishments for me don't become less impressive because players needed longer to accomplish them or had worse aligulac rating during them, worse winrates etc.
|
On June 01 2025 16:16 Mizenhauer wrote:
My only question is, given that Rogue's GOAT score is so disproportionately low compared to the other players in this study, do you feel it's an indication of some fundamental error in your process? I say this because that kind of skew in results leads me to believe that something is inherently incorrect when it comes to the calculation/tabulation of your GOAT score. Maybe that's ChatGPT (it's pretty damning).
100% agree. This way of thinking about GOATs inevitably leads to such bizarre results. It’s the kind of process that for sure leaves Muhammad Ali no where near the top 10 on a boxing GOAT list (that’s not conjecture, stats-based boxing rankings have consistently produced this result). Basically a deep, conceptual misunderstanding of what “greatness” means.
It’s interesting that this is framed as having taken seriously critiques of the previous analysis when it ironically reproduces this conceptual error so glaringly and fails to even attempt to address it.
|
On June 02 2025 04:00 PremoBeats wrote:Show nested quote +On June 02 2025 02:18 JJH777 wrote:On June 02 2025 01:06 PremoBeats wrote:On June 01 2025 23:54 Hider wrote:On June 01 2025 16:16 Mizenhauer wrote:
My only question is, given that Rogue's GOAT score is so disproportionately low compared to the other players in this study, do you feel it's an indication of some fundamental error in your process? I say this because that kind of skew in results leads me to believe that something is inherently incorrect when it comes to the calculation/tabulation of your GOAT score. Maybe that's ChatGPT (it's pretty damning). Isn't the thing about Rogue that he was more streaky or perhaps more clutchy. I never looked at Rogue and thought he was dominant or even the best player in the world at any point in time. Personally I put MVP/Inno over Rogue on my GOAT list. A criteria for being the GOAT is imo that you were the best player in the world for some period in time. I weight MVP's era of dominance and the early KESPA era multiple times higher per tournament than what happened in the past few years. On the other hand we had more tournaments back then so that offsets it partially. On a per tournament basis I would give perhaps 5 times more weight to a tournament during Starcraft's peak era than today. However, the margin of dominance also matters, and that's where Serral truly shines and still comes in at number 1. MVP had perhaps 6 months where he looked a level above everyone else, but after that it felt more like he was one of the better players but just pretty clutchy. In contrast Serral has always been the favorite or huge favorite for 7 years straight. 5 times more? Like a blank multiplier? Even though there were Premier Tournaments where two to three top tier Koreans flew to foreign regions to farm them? These situations are comparable to Serral winning in Europe and you'd want to award these tournaments 5 times the amount in relation to modern ones? I don't think that this is a sound multiplier, unless I misunderstood you. On June 02 2025 00:30 JJH777 wrote:
Aligulac being 20% of your determination is insane. Aligulac is a bad ELO system for a bunch of reasons. It consistently overrates people from outside of the KR scene. There are tons of examples even before Serral was notable. Serral specifically hit rank 1 for the first time when he still had <50% win rate vs Koreans. For much of his career his ranking was inflated by playing a huge amount of games vs mid to low tier EU players which he always swept. Maru would have swept those too. If you put Maru into 3 8 player round robin groups vs non Serral/Reynor/Clem EU players per year (format of EU regional for most of the years being discussed) what do you think happens to his aligulac rating? I guarantee it skyrockets. It's also funny that you punish Marus tournament winrrate for winning against "low tier" Koreans but don't take that into account that Serral's aligulac ranking heavily benefits from beating up low tier EU players.
I also heavily disagree with the idea of average placement/efficiency/percentage of won tournaments in all premiers played being important. That just specifically punishes players with long careers who consistently qualified for premiers. A longer career with consistent premier qualifications should be a boon to goat candidacy not a negative. The fact that Marus ranking in these categories would go up dramatically if you exclude his pre 2018 results shows alone how flawed it is. Maru being an S tier player from 2013-2018 should not penalize his chances of being goat.
I know for some of these aspects you slightly adjusted for this by determining prime years but I disagree with it being a consideration at all. Certainly not a combined 42.5% of your ranking. 1. The overall result wouldn't change (Serral would still lead with nearly 300 points), even if I removed the Aligulac-metric entirely 2. I think you got this the wrong way round. The machine's algorithm saw the explosion of wins Serral had versus Europeans and through cross-regional comparisons correctly "predicted" that he would defeat the Koreans as well, which was exactly what happened in the first quarter of 2018 when his win rates skyrocketed to over 80%. What you mentioned is a super rare case, that shows how well Aligulac works cross-regionally. These inflation-arguments were a thing in the early 2010s... not anymore in 2017. Further, after 2018, Serral played enough Koreans per year, which made the calculation pretty reliable. 3. I only pointed the Korean's inflation against Serral through weaker Korean players out. As I wrote in the article, I did not actually incorporate these findings in the calculation. 4. No, it punishes no one. It simply shows, that there are some players that have the ability to always finish on top and others that don't. That the player you favor by chance is not good in this metric, does not mean that the metric per se is bad. 5. Some argue rightfully, that it would be unfair to players like Rain and Mvp who had short, kick-started careers if one discounts the worse years of players, hence I went for a mix. In the tournament score Maru's longer careers ears him more points. And if he was as efficient as Serral over 7 years, he'd easily be up there with him. That is simply what a fair analysis shows... strengths and weaknesses. Aligulac called King Kong a top 10 player in the world as late as 2014. I'm pretty sure he has never beaten a top pro. He got that ranking by being dominant in his region while never having any global results. He stayed in the top 50 range for years despite the fact that on a skill basis he was likely never even top 100. Aligulac has huge issues judging players who dominate their own region. Another good example is Polt. Polt is in the top 15 all time on aligulac. He was regularly top 5 and even 1st while active. He did that by dominating NA. He was never truly competitive with top GSL players post 2011 but dominating a weaker region kept him at the top of aligulac for ages. Saying aligulac was just predicting Serral's future success is a copout when that future success wasn't until over a year after he hit rank 1. During which time he regularly lost matches to Koreans but still during that entire period he was almost always 1st with just a few falls down to 2-3. I'm definitely not saying he didn't deserve it the majority of the time he had it but the margin he had it by was heavily inflated and there were many times he would have been passed with a system that didn't over reward winning against far weaker competition. I don't know how possible it is to calculate, probably depends on how accessible aligulac's source code/algorithm and result sets are but I am genuinely curious about how aligulac would have changed if Maru played 24 bo3s vs mid and lower rank EU players (basically everyone sans Serral and Reynor until 2022 then everyone besides them and Clem after that) every year from 2017-2024. Giving him whatever his current winrrate is vs that group in those bo3s. Based on aligulac's history of over rewarding beating lower competition I am confident his rating would increase significantly but I can't say for sure. I may try to do that someday when I have free time. Marus "ability to finish on top" is punished by playing in an era with far more active full time pros and by having a far higher amount of events played in. Those criteria definitely uniquely punishes Marus career as it is the longest of any top pro by a considerable margin. By those criteria he would literally rank higher if he had just failed to qualify for events than if he qualifies for them. The fact that you can't agree that is backwards is crazy to me. Serral's metrics in all categories besides tournament results are inflated by not playing in GSL. If Serral played in GSL regularly it's extremely likely his aligulac, matchup win rates, efficiency, average tournament placement, and percentage of won tournaments would all decline. I do agree he would likely win some GSLs which would benefit his tournament score but when you're only giving tournament score 22% of the overall weight it's not like that makes much difference. - I already said that Aligulac had issues in the 2010s. - Didn't he hit rank 1 on the last list of 2017? And started winning against Koreans in January 2018? I mean.. that is exactly what the machine is built to do. Weigh players relative to their statistics against each other. - Which exact time frame are you talking about when talking about Maru? And which metric? - Why would Serral lose more games in GSL than versus top Koreans at other Premier Tournaments? Because of its structure?
You said early 2010s. The issues I mentioned were 2014-2017. A more recent example would be heromarine shortly after the open cups started. He won a bunch in a row because most top players didn't play in them and shot up the rankings and was top 10 for a while. I like heromarine but I don't really believe he was ever top 10 worldwide in terms of skill. Even after his one Kato top 4 he didn't follow that up with any major results suggesting it was a fluke.
Regardless even 2014 is late enough to show issues. Those first years were the most active years of SC2. Why would a system that by your own admission didn't work with 100,000 data points suddenly work with 200,000-300,000? Numbers are estimated based on aligulac's search function.
The timeframe for Maru would be 2012-2017 and the metrics I already specified were efficiency, average placement, and percentage of won tournaments. The way you measure those Maru would have been better off being an unknown/inactive player who never qualified for anything during that era because then his 2018 forward timeframe would produce much better results in those metrics. Any measuring system that determines being inactive superior to Maru's 2012-2017 results is crazy.
Prep format, offline in Korea, having to travel back and forth to events and simply a higher sample size. Some of Serral's impressive win rate years vs Koreans are based on a very small number of series. For example 2018 I doubt he could have kept that winrrate as high if he played more than 28 series.
|
Northern Ireland25035 Posts
On June 02 2025 08:39 JJH777 wrote:Show nested quote +On June 02 2025 04:00 PremoBeats wrote:On June 02 2025 02:18 JJH777 wrote:On June 02 2025 01:06 PremoBeats wrote:On June 01 2025 23:54 Hider wrote:On June 01 2025 16:16 Mizenhauer wrote:
My only question is, given that Rogue's GOAT score is so disproportionately low compared to the other players in this study, do you feel it's an indication of some fundamental error in your process? I say this because that kind of skew in results leads me to believe that something is inherently incorrect when it comes to the calculation/tabulation of your GOAT score. Maybe that's ChatGPT (it's pretty damning). Isn't the thing about Rogue that he was more streaky or perhaps more clutchy. I never looked at Rogue and thought he was dominant or even the best player in the world at any point in time. Personally I put MVP/Inno over Rogue on my GOAT list. A criteria for being the GOAT is imo that you were the best player in the world for some period in time. I weight MVP's era of dominance and the early KESPA era multiple times higher per tournament than what happened in the past few years. On the other hand we had more tournaments back then so that offsets it partially. On a per tournament basis I would give perhaps 5 times more weight to a tournament during Starcraft's peak era than today. However, the margin of dominance also matters, and that's where Serral truly shines and still comes in at number 1. MVP had perhaps 6 months where he looked a level above everyone else, but after that it felt more like he was one of the better players but just pretty clutchy. In contrast Serral has always been the favorite or huge favorite for 7 years straight. 5 times more? Like a blank multiplier? Even though there were Premier Tournaments where two to three top tier Koreans flew to foreign regions to farm them? These situations are comparable to Serral winning in Europe and you'd want to award these tournaments 5 times the amount in relation to modern ones? I don't think that this is a sound multiplier, unless I misunderstood you. On June 02 2025 00:30 JJH777 wrote:
Aligulac being 20% of your determination is insane. Aligulac is a bad ELO system for a bunch of reasons. It consistently overrates people from outside of the KR scene. There are tons of examples even before Serral was notable. Serral specifically hit rank 1 for the first time when he still had <50% win rate vs Koreans. For much of his career his ranking was inflated by playing a huge amount of games vs mid to low tier EU players which he always swept. Maru would have swept those too. If you put Maru into 3 8 player round robin groups vs non Serral/Reynor/Clem EU players per year (format of EU regional for most of the years being discussed) what do you think happens to his aligulac rating? I guarantee it skyrockets. It's also funny that you punish Marus tournament winrrate for winning against "low tier" Koreans but don't take that into account that Serral's aligulac ranking heavily benefits from beating up low tier EU players.
I also heavily disagree with the idea of average placement/efficiency/percentage of won tournaments in all premiers played being important. That just specifically punishes players with long careers who consistently qualified for premiers. A longer career with consistent premier qualifications should be a boon to goat candidacy not a negative. The fact that Marus ranking in these categories would go up dramatically if you exclude his pre 2018 results shows alone how flawed it is. Maru being an S tier player from 2013-2018 should not penalize his chances of being goat.
I know for some of these aspects you slightly adjusted for this by determining prime years but I disagree with it being a consideration at all. Certainly not a combined 42.5% of your ranking. 1. The overall result wouldn't change (Serral would still lead with nearly 300 points), even if I removed the Aligulac-metric entirely 2. I think you got this the wrong way round. The machine's algorithm saw the explosion of wins Serral had versus Europeans and through cross-regional comparisons correctly "predicted" that he would defeat the Koreans as well, which was exactly what happened in the first quarter of 2018 when his win rates skyrocketed to over 80%. What you mentioned is a super rare case, that shows how well Aligulac works cross-regionally. These inflation-arguments were a thing in the early 2010s... not anymore in 2017. Further, after 2018, Serral played enough Koreans per year, which made the calculation pretty reliable. 3. I only pointed the Korean's inflation against Serral through weaker Korean players out. As I wrote in the article, I did not actually incorporate these findings in the calculation. 4. No, it punishes no one. It simply shows, that there are some players that have the ability to always finish on top and others that don't. That the player you favor by chance is not good in this metric, does not mean that the metric per se is bad. 5. Some argue rightfully, that it would be unfair to players like Rain and Mvp who had short, kick-started careers if one discounts the worse years of players, hence I went for a mix. In the tournament score Maru's longer careers ears him more points. And if he was as efficient as Serral over 7 years, he'd easily be up there with him. That is simply what a fair analysis shows... strengths and weaknesses. Aligulac called King Kong a top 10 player in the world as late as 2014. I'm pretty sure he has never beaten a top pro. He got that ranking by being dominant in his region while never having any global results. He stayed in the top 50 range for years despite the fact that on a skill basis he was likely never even top 100. Aligulac has huge issues judging players who dominate their own region. Another good example is Polt. Polt is in the top 15 all time on aligulac. He was regularly top 5 and even 1st while active. He did that by dominating NA. He was never truly competitive with top GSL players post 2011 but dominating a weaker region kept him at the top of aligulac for ages. Saying aligulac was just predicting Serral's future success is a copout when that future success wasn't until over a year after he hit rank 1. During which time he regularly lost matches to Koreans but still during that entire period he was almost always 1st with just a few falls down to 2-3. I'm definitely not saying he didn't deserve it the majority of the time he had it but the margin he had it by was heavily inflated and there were many times he would have been passed with a system that didn't over reward winning against far weaker competition. I don't know how possible it is to calculate, probably depends on how accessible aligulac's source code/algorithm and result sets are but I am genuinely curious about how aligulac would have changed if Maru played 24 bo3s vs mid and lower rank EU players (basically everyone sans Serral and Reynor until 2022 then everyone besides them and Clem after that) every year from 2017-2024. Giving him whatever his current winrrate is vs that group in those bo3s. Based on aligulac's history of over rewarding beating lower competition I am confident his rating would increase significantly but I can't say for sure. I may try to do that someday when I have free time. Marus "ability to finish on top" is punished by playing in an era with far more active full time pros and by having a far higher amount of events played in. Those criteria definitely uniquely punishes Marus career as it is the longest of any top pro by a considerable margin. By those criteria he would literally rank higher if he had just failed to qualify for events than if he qualifies for them. The fact that you can't agree that is backwards is crazy to me. Serral's metrics in all categories besides tournament results are inflated by not playing in GSL. If Serral played in GSL regularly it's extremely likely his aligulac, matchup win rates, efficiency, average tournament placement, and percentage of won tournaments would all decline. I do agree he would likely win some GSLs which would benefit his tournament score but when you're only giving tournament score 22% of the overall weight it's not like that makes much difference. - I already said that Aligulac had issues in the 2010s. - Didn't he hit rank 1 on the last list of 2017? And started winning against Koreans in January 2018? I mean.. that is exactly what the machine is built to do. Weigh players relative to their statistics against each other. - Which exact time frame are you talking about when talking about Maru? And which metric? - Why would Serral lose more games in GSL than versus top Koreans at other Premier Tournaments? Because of its structure? You said early 2010s. The issues I mentioned were 2014-2017. A more recent example would be heromarine shortly after the open cups started. He won a bunch in a row because most top players didn't play in them and shot up the rankings and was top 10 for a while. I like heromarine but I don't really believe he was ever top 10 worldwide in terms of skill. Even after his one Kato top 4 he didn't follow that up with any major results suggesting it was a fluke. Regardless even 2014 is late enough to show issues. Those first years were the most active years of SC2. Why would a system that by your own admission didn't work with 100,000 data points suddenly work with 200,000-300,000? Numbers are estimated based on aligulac's search function. The timeframe for Maru would be 2012-2017 and the metrics I already specified were efficiency, average placement, and percentage of won tournaments. The way you measure those Maru would have been better off being an unknown/inactive player who never qualified for anything during that era because then his 2018 forward timeframe would produce much better results in those metrics. Any measuring system that determines being inactive superior to Maru's 2012-2017 results is crazy. Prep format, offline in Korea, having to travel back and forth to events and simply a higher sample size. Some of Serral's impressive win rate years vs Koreans are based on a very small number of series. For example 2018 I doubt he could have kept that winrrate as high if he played more than 28 series.
28 series isn’t a small amount of series at all. It’s enough to win 4 GSLs if you sweep the board and don’t drop one.
Looking at his victims that year, it’s also a veritable who’s-who of the crème de la crème of that time from Korea.
I could see it dropping if he only played those same pool of opponents over and over again.
On the flipside, it’s not even a versus Korean win rate really, it’s almost a win rate specifically against top, top tier Koreans. I think there’s a case that the form he was in, 2018 Serral could have raised his win rate if he was unleashed on a bunch of perpetual GSL Ro32/Ro16 at best level Koreans
In some subsequent years I think that’s less the case, in 2018 the opposition really was somewhat filtered that way.
I’m not sure you take Serral’s opponents that year, have another Korean play those sets and have any of them get that win rate.
He has quite a few years that are streets ahead of most by that specific metric, and he’s rarely playing the lower tier Koreans.
|
Northern Ireland25035 Posts
@Premo nice one, great job, should stir some debate anyway! Saw it on Reddit too said he’ll there
I appreciate you showing the raw numbers alongside your adjustments too. Makes things less opaque. It irks me when people show their adjusted scores without the raw data, it makes it hard to sorta judge at a glance, as well as assess the weightings themselves.
I think you can’t win some battles, some crit is fine, some is a bit harsh.
I think factoring in ELO, good idea. People bring up Aligulac’s flaws. I also agree it has flaws. But there isn’t an alternative ELO ranking set to use, so what do?
I’d rather you just threw it in, it’s some interesting data anyway, one can take it or leave it.
People will claim x era is weaker or stronger, but then if you weight for it, ‘hey you can’t do that that’s arbitrary!’
Overall I think it’s a great read, and I think a reader can take what they want from it, bar hostility which I think is unwarranted. Worst case it’s some numbers someone has pulled, saves me going through Aligulac! I skimmed it mostly just looking at the numbers and some of your rationales but I’ll return to really pore over the whole thing properly.
We cycle back to my usual point on this. And I’m referring to fans of particular players or positions, not people who write things like this.
You can either say x is the GOAT because various things I value and intangibles, or do something in the realms of cold hard numbers. Or blend the two to some degree.
You can’t just pick and choose which to ignore, or highlight when it suits. You know like player x > y because y doesn’t have a particular thing ticked off. But for player z them not having a particular thing doesn’t matter.
Thus Rogue cannot be the GOAT IMO :p To take one example.
I kid, I think he can, but it has to be a claim heavily based on the intangible, things like his clutchness, his fiendish plays at times etc.
That’s absolutely A-OK by me. I still have Inno up at #3 for me, many, many disagree with that. But the peak Machine was the most terrifying player, relative to the level of his peers that I’ve ever seen, and I imagine we will see. Wasn’t long til folks caught up, but it’s was like watch a semi-pro like mid-GM against a good pro now. It wasn’t that he was better, it was like another level.
If people want to do that for Rogue too, that’s bloody swell if you ask me. But just do that
Once people try to do it by qualifiers and caveats it becomes a mess.
You end up with absurd situations where like someone will have Inno > Serral because he won most of his titles in the Kespa era, but also Rogue > Serral because he’s got Starleagues to tiebreak their WCs, but also Rogue > Maru despite Maru having more Starleagues because Rogue’s got WCs, but also Rogue > sOs because he won more titles overall.
Now, to clarify I don’t mind people using qualifiers whatsoever, I’m taking about sorta chaining them up and applying them in different head-to-heads differently
|
On June 02 2025 09:21 WombaT wrote:Show nested quote +On June 02 2025 08:39 JJH777 wrote:On June 02 2025 04:00 PremoBeats wrote:On June 02 2025 02:18 JJH777 wrote:On June 02 2025 01:06 PremoBeats wrote:On June 01 2025 23:54 Hider wrote:On June 01 2025 16:16 Mizenhauer wrote:
My only question is, given that Rogue's GOAT score is so disproportionately low compared to the other players in this study, do you feel it's an indication of some fundamental error in your process? I say this because that kind of skew in results leads me to believe that something is inherently incorrect when it comes to the calculation/tabulation of your GOAT score. Maybe that's ChatGPT (it's pretty damning). Isn't the thing about Rogue that he was more streaky or perhaps more clutchy. I never looked at Rogue and thought he was dominant or even the best player in the world at any point in time. Personally I put MVP/Inno over Rogue on my GOAT list. A criteria for being the GOAT is imo that you were the best player in the world for some period in time. I weight MVP's era of dominance and the early KESPA era multiple times higher per tournament than what happened in the past few years. On the other hand we had more tournaments back then so that offsets it partially. On a per tournament basis I would give perhaps 5 times more weight to a tournament during Starcraft's peak era than today. However, the margin of dominance also matters, and that's where Serral truly shines and still comes in at number 1. MVP had perhaps 6 months where he looked a level above everyone else, but after that it felt more like he was one of the better players but just pretty clutchy. In contrast Serral has always been the favorite or huge favorite for 7 years straight. 5 times more? Like a blank multiplier? Even though there were Premier Tournaments where two to three top tier Koreans flew to foreign regions to farm them? These situations are comparable to Serral winning in Europe and you'd want to award these tournaments 5 times the amount in relation to modern ones? I don't think that this is a sound multiplier, unless I misunderstood you. On June 02 2025 00:30 JJH777 wrote:
Aligulac being 20% of your determination is insane. Aligulac is a bad ELO system for a bunch of reasons. It consistently overrates people from outside of the KR scene. There are tons of examples even before Serral was notable. Serral specifically hit rank 1 for the first time when he still had <50% win rate vs Koreans. For much of his career his ranking was inflated by playing a huge amount of games vs mid to low tier EU players which he always swept. Maru would have swept those too. If you put Maru into 3 8 player round robin groups vs non Serral/Reynor/Clem EU players per year (format of EU regional for most of the years being discussed) what do you think happens to his aligulac rating? I guarantee it skyrockets. It's also funny that you punish Marus tournament winrrate for winning against "low tier" Koreans but don't take that into account that Serral's aligulac ranking heavily benefits from beating up low tier EU players.
I also heavily disagree with the idea of average placement/efficiency/percentage of won tournaments in all premiers played being important. That just specifically punishes players with long careers who consistently qualified for premiers. A longer career with consistent premier qualifications should be a boon to goat candidacy not a negative. The fact that Marus ranking in these categories would go up dramatically if you exclude his pre 2018 results shows alone how flawed it is. Maru being an S tier player from 2013-2018 should not penalize his chances of being goat.
I know for some of these aspects you slightly adjusted for this by determining prime years but I disagree with it being a consideration at all. Certainly not a combined 42.5% of your ranking. 1. The overall result wouldn't change (Serral would still lead with nearly 300 points), even if I removed the Aligulac-metric entirely 2. I think you got this the wrong way round. The machine's algorithm saw the explosion of wins Serral had versus Europeans and through cross-regional comparisons correctly "predicted" that he would defeat the Koreans as well, which was exactly what happened in the first quarter of 2018 when his win rates skyrocketed to over 80%. What you mentioned is a super rare case, that shows how well Aligulac works cross-regionally. These inflation-arguments were a thing in the early 2010s... not anymore in 2017. Further, after 2018, Serral played enough Koreans per year, which made the calculation pretty reliable. 3. I only pointed the Korean's inflation against Serral through weaker Korean players out. As I wrote in the article, I did not actually incorporate these findings in the calculation. 4. No, it punishes no one. It simply shows, that there are some players that have the ability to always finish on top and others that don't. That the player you favor by chance is not good in this metric, does not mean that the metric per se is bad. 5. Some argue rightfully, that it would be unfair to players like Rain and Mvp who had short, kick-started careers if one discounts the worse years of players, hence I went for a mix. In the tournament score Maru's longer careers ears him more points. And if he was as efficient as Serral over 7 years, he'd easily be up there with him. That is simply what a fair analysis shows... strengths and weaknesses. Aligulac called King Kong a top 10 player in the world as late as 2014. I'm pretty sure he has never beaten a top pro. He got that ranking by being dominant in his region while never having any global results. He stayed in the top 50 range for years despite the fact that on a skill basis he was likely never even top 100. Aligulac has huge issues judging players who dominate their own region. Another good example is Polt. Polt is in the top 15 all time on aligulac. He was regularly top 5 and even 1st while active. He did that by dominating NA. He was never truly competitive with top GSL players post 2011 but dominating a weaker region kept him at the top of aligulac for ages. Saying aligulac was just predicting Serral's future success is a copout when that future success wasn't until over a year after he hit rank 1. During which time he regularly lost matches to Koreans but still during that entire period he was almost always 1st with just a few falls down to 2-3. I'm definitely not saying he didn't deserve it the majority of the time he had it but the margin he had it by was heavily inflated and there were many times he would have been passed with a system that didn't over reward winning against far weaker competition. I don't know how possible it is to calculate, probably depends on how accessible aligulac's source code/algorithm and result sets are but I am genuinely curious about how aligulac would have changed if Maru played 24 bo3s vs mid and lower rank EU players (basically everyone sans Serral and Reynor until 2022 then everyone besides them and Clem after that) every year from 2017-2024. Giving him whatever his current winrrate is vs that group in those bo3s. Based on aligulac's history of over rewarding beating lower competition I am confident his rating would increase significantly but I can't say for sure. I may try to do that someday when I have free time. Marus "ability to finish on top" is punished by playing in an era with far more active full time pros and by having a far higher amount of events played in. Those criteria definitely uniquely punishes Marus career as it is the longest of any top pro by a considerable margin. By those criteria he would literally rank higher if he had just failed to qualify for events than if he qualifies for them. The fact that you can't agree that is backwards is crazy to me. Serral's metrics in all categories besides tournament results are inflated by not playing in GSL. If Serral played in GSL regularly it's extremely likely his aligulac, matchup win rates, efficiency, average tournament placement, and percentage of won tournaments would all decline. I do agree he would likely win some GSLs which would benefit his tournament score but when you're only giving tournament score 22% of the overall weight it's not like that makes much difference. - I already said that Aligulac had issues in the 2010s. - Didn't he hit rank 1 on the last list of 2017? And started winning against Koreans in January 2018? I mean.. that is exactly what the machine is built to do. Weigh players relative to their statistics against each other. - Which exact time frame are you talking about when talking about Maru? And which metric? - Why would Serral lose more games in GSL than versus top Koreans at other Premier Tournaments? Because of its structure? You said early 2010s. The issues I mentioned were 2014-2017. A more recent example would be heromarine shortly after the open cups started. He won a bunch in a row because most top players didn't play in them and shot up the rankings and was top 10 for a while. I like heromarine but I don't really believe he was ever top 10 worldwide in terms of skill. Even after his one Kato top 4 he didn't follow that up with any major results suggesting it was a fluke. Regardless even 2014 is late enough to show issues. Those first years were the most active years of SC2. Why would a system that by your own admission didn't work with 100,000 data points suddenly work with 200,000-300,000? Numbers are estimated based on aligulac's search function. The timeframe for Maru would be 2012-2017 and the metrics I already specified were efficiency, average placement, and percentage of won tournaments. The way you measure those Maru would have been better off being an unknown/inactive player who never qualified for anything during that era because then his 2018 forward timeframe would produce much better results in those metrics. Any measuring system that determines being inactive superior to Maru's 2012-2017 results is crazy. Prep format, offline in Korea, having to travel back and forth to events and simply a higher sample size. Some of Serral's impressive win rate years vs Koreans are based on a very small number of series. For example 2018 I doubt he could have kept that winrrate as high if he played more than 28 series. On the flipside, it’s not even a versus Korean win rate really, it’s almost a win rate specifically against top, top tier Koreans. I think there’s a case that the form he was in, 2018 Serral could have raised his win rate if he was unleashed on a bunch of perpetual GSL Ro32/Ro16 at best level Koreans In some subsequent years I think that’s less the case, in 2018 the opposition really was somewhat filtered that way. I’m not sure you take Serral’s opponents that year, have another Korean play those sets and have any of them get that win rate. He has quite a few years that are streets ahead of most by that specific metric, and he’s rarely playing the lower tier Koreans.
What ifs are always tricky, but I'm not sold he'd increase his winrate if he had played in gsl that year. It's just likely he would've been given the reynor treatment and kicked out in groups. For all the talk about his good winrate against koreans, in 2018 his matches against koreans (in his best case scenario of no gsl level prep) were competitive but he wasn't dominating them.
Iem 2018: went 3-2 against trap in the ro8 and 0-3 against classic in the semis
Wesg 2018: went 0-3 against maru in the semis
His best result by far that whole year was blizzcon, and even then stats took 2 games off of him.
|
Northern Ireland25035 Posts
On June 02 2025 04:26 Charoisaur wrote:Show nested quote +Most of the metrics can show both, although some are harder to sustain for long periods, I agree. Yeah, that's the point. It just heavily punishes players who have high longevity but were never really dominant (like Dark), compared to players who only shined during a brief period but were very dominant (like Mvp). I don't see any objective reasons why the achievements of someone like Dark should be worth less, just because he needed more time to accomplish them. In the opposite, he proved he can succeed in many more different metas and fields of competition than Mvp, which should count for something. It's a good ranking for most dominant player of all time, but at least for me that's not (solely) what Greatest of all time means.
I think people sleep on Dark a lot anyway myself, I actually think he’d probably score quite well with these metrics anyway too, not that I wish Premo more Excel pain!
It doesn’t necessarily punish players for longevity, they’re punished for not being the best for a span versus someone who might have been, for me anyway.
For sake of argument, and I’m stripping Rain of a Starleague to equalise. I’m a huge, huge fan of Gumigod. Is anyone not tbh, aside from his opponents when he’s doing something funky? Anyway yeah, they’ve both now got a single GSL. Which would you pick of Rain or Gumigod first in a GOAT list? I think the speed you accomplish things does matter, to some degree. Although really I only factor it in if player x left the scene prematurely
I can’t really think of many sports for example, where names frequently put into the GOAT hat weren’t dominant. In individual sports they gotta win big, in team sports they may just be excelling individually, winning a lot of trophies, or both.
Boxing is probably one, but that’s super specific to the way the sport is run (i.e. total bullshit often).
I think as well, SC2’s truly dominant players all have some other X factor as well, those greatness intangibles.
Mvp - Ahead of the curve in a new game. He put the flesh on the bones of how Terran played, which is for me pretty underrated for a strategy game. Most others were merely standing on the shoulders of giants and iterating. Also Old Man MVP’s last silver, for me maybe more ‘great’ than his wins, battling against the ravages of injury, and scraping by with his wits.
Innovation - Raised the bar, very quickly. Faster than his Kespa cohorts could match, and ultimately higher than many of the previous pros could keep up with. For a relatively brief period he was playing StarCraft at a level we’d simply not seen.
Serral - The first foreigner to break the glass ceiling proper. And he didn’t just break it, he leapt so high he punched through the next couple. He didn’t just peak to get that WC for the foreigners with a miracle, TIME-style run, this dude was now the guy to beat. Absurd consistency and shattered basically every stat record going.
Maru - I mean it’s Maru. He’s, at some point mastered every facet of the game, from cheeky cheese, to brutal pushes and timings, to unparalleled levels of defensive late-game slugfests. ‘Wait, other players use vikings to counter colossus?’ Perhaps Clem may one day catch up, but if there’s a player with the best micro moments highlight reel it’s Maru and it’s not even particularly close.
Voldemort - Incredible player at an incredibly young age. We’d seen many good ones, we hadn’t really seen one, however briefly be the best.
Rogue - His tournament goals are not to win it, they are to reach the final because if he does that, he’s already won.
Anyway, not to ramble too much. I rate Dark super high as well, and I think you make a great point about him adapting not just to metas, but across games. I do have to say I think a period of dominance really does matter. Not just for its own sake but it’s usually accompanied by some other facet of greatness.
|
Northern Ireland25035 Posts
On June 02 2025 11:15 lokol4890 wrote:Show nested quote +On June 02 2025 09:21 WombaT wrote:On June 02 2025 08:39 JJH777 wrote:On June 02 2025 04:00 PremoBeats wrote:On June 02 2025 02:18 JJH777 wrote:On June 02 2025 01:06 PremoBeats wrote:On June 01 2025 23:54 Hider wrote:On June 01 2025 16:16 Mizenhauer wrote:
My only question is, given that Rogue's GOAT score is so disproportionately low compared to the other players in this study, do you feel it's an indication of some fundamental error in your process? I say this because that kind of skew in results leads me to believe that something is inherently incorrect when it comes to the calculation/tabulation of your GOAT score. Maybe that's ChatGPT (it's pretty damning). Isn't the thing about Rogue that he was more streaky or perhaps more clutchy. I never looked at Rogue and thought he was dominant or even the best player in the world at any point in time. Personally I put MVP/Inno over Rogue on my GOAT list. A criteria for being the GOAT is imo that you were the best player in the world for some period in time. I weight MVP's era of dominance and the early KESPA era multiple times higher per tournament than what happened in the past few years. On the other hand we had more tournaments back then so that offsets it partially. On a per tournament basis I would give perhaps 5 times more weight to a tournament during Starcraft's peak era than today. However, the margin of dominance also matters, and that's where Serral truly shines and still comes in at number 1. MVP had perhaps 6 months where he looked a level above everyone else, but after that it felt more like he was one of the better players but just pretty clutchy. In contrast Serral has always been the favorite or huge favorite for 7 years straight. 5 times more? Like a blank multiplier? Even though there were Premier Tournaments where two to three top tier Koreans flew to foreign regions to farm them? These situations are comparable to Serral winning in Europe and you'd want to award these tournaments 5 times the amount in relation to modern ones? I don't think that this is a sound multiplier, unless I misunderstood you. On June 02 2025 00:30 JJH777 wrote:
Aligulac being 20% of your determination is insane. Aligulac is a bad ELO system for a bunch of reasons. It consistently overrates people from outside of the KR scene. There are tons of examples even before Serral was notable. Serral specifically hit rank 1 for the first time when he still had <50% win rate vs Koreans. For much of his career his ranking was inflated by playing a huge amount of games vs mid to low tier EU players which he always swept. Maru would have swept those too. If you put Maru into 3 8 player round robin groups vs non Serral/Reynor/Clem EU players per year (format of EU regional for most of the years being discussed) what do you think happens to his aligulac rating? I guarantee it skyrockets. It's also funny that you punish Marus tournament winrrate for winning against "low tier" Koreans but don't take that into account that Serral's aligulac ranking heavily benefits from beating up low tier EU players.
I also heavily disagree with the idea of average placement/efficiency/percentage of won tournaments in all premiers played being important. That just specifically punishes players with long careers who consistently qualified for premiers. A longer career with consistent premier qualifications should be a boon to goat candidacy not a negative. The fact that Marus ranking in these categories would go up dramatically if you exclude his pre 2018 results shows alone how flawed it is. Maru being an S tier player from 2013-2018 should not penalize his chances of being goat.
I know for some of these aspects you slightly adjusted for this by determining prime years but I disagree with it being a consideration at all. Certainly not a combined 42.5% of your ranking. 1. The overall result wouldn't change (Serral would still lead with nearly 300 points), even if I removed the Aligulac-metric entirely 2. I think you got this the wrong way round. The machine's algorithm saw the explosion of wins Serral had versus Europeans and through cross-regional comparisons correctly "predicted" that he would defeat the Koreans as well, which was exactly what happened in the first quarter of 2018 when his win rates skyrocketed to over 80%. What you mentioned is a super rare case, that shows how well Aligulac works cross-regionally. These inflation-arguments were a thing in the early 2010s... not anymore in 2017. Further, after 2018, Serral played enough Koreans per year, which made the calculation pretty reliable. 3. I only pointed the Korean's inflation against Serral through weaker Korean players out. As I wrote in the article, I did not actually incorporate these findings in the calculation. 4. No, it punishes no one. It simply shows, that there are some players that have the ability to always finish on top and others that don't. That the player you favor by chance is not good in this metric, does not mean that the metric per se is bad. 5. Some argue rightfully, that it would be unfair to players like Rain and Mvp who had short, kick-started careers if one discounts the worse years of players, hence I went for a mix. In the tournament score Maru's longer careers ears him more points. And if he was as efficient as Serral over 7 years, he'd easily be up there with him. That is simply what a fair analysis shows... strengths and weaknesses. Aligulac called King Kong a top 10 player in the world as late as 2014. I'm pretty sure he has never beaten a top pro. He got that ranking by being dominant in his region while never having any global results. He stayed in the top 50 range for years despite the fact that on a skill basis he was likely never even top 100. Aligulac has huge issues judging players who dominate their own region. Another good example is Polt. Polt is in the top 15 all time on aligulac. He was regularly top 5 and even 1st while active. He did that by dominating NA. He was never truly competitive with top GSL players post 2011 but dominating a weaker region kept him at the top of aligulac for ages. Saying aligulac was just predicting Serral's future success is a copout when that future success wasn't until over a year after he hit rank 1. During which time he regularly lost matches to Koreans but still during that entire period he was almost always 1st with just a few falls down to 2-3. I'm definitely not saying he didn't deserve it the majority of the time he had it but the margin he had it by was heavily inflated and there were many times he would have been passed with a system that didn't over reward winning against far weaker competition. I don't know how possible it is to calculate, probably depends on how accessible aligulac's source code/algorithm and result sets are but I am genuinely curious about how aligulac would have changed if Maru played 24 bo3s vs mid and lower rank EU players (basically everyone sans Serral and Reynor until 2022 then everyone besides them and Clem after that) every year from 2017-2024. Giving him whatever his current winrrate is vs that group in those bo3s. Based on aligulac's history of over rewarding beating lower competition I am confident his rating would increase significantly but I can't say for sure. I may try to do that someday when I have free time. Marus "ability to finish on top" is punished by playing in an era with far more active full time pros and by having a far higher amount of events played in. Those criteria definitely uniquely punishes Marus career as it is the longest of any top pro by a considerable margin. By those criteria he would literally rank higher if he had just failed to qualify for events than if he qualifies for them. The fact that you can't agree that is backwards is crazy to me. Serral's metrics in all categories besides tournament results are inflated by not playing in GSL. If Serral played in GSL regularly it's extremely likely his aligulac, matchup win rates, efficiency, average tournament placement, and percentage of won tournaments would all decline. I do agree he would likely win some GSLs which would benefit his tournament score but when you're only giving tournament score 22% of the overall weight it's not like that makes much difference. - I already said that Aligulac had issues in the 2010s. - Didn't he hit rank 1 on the last list of 2017? And started winning against Koreans in January 2018? I mean.. that is exactly what the machine is built to do. Weigh players relative to their statistics against each other. - Which exact time frame are you talking about when talking about Maru? And which metric? - Why would Serral lose more games in GSL than versus top Koreans at other Premier Tournaments? Because of its structure? You said early 2010s. The issues I mentioned were 2014-2017. A more recent example would be heromarine shortly after the open cups started. He won a bunch in a row because most top players didn't play in them and shot up the rankings and was top 10 for a while. I like heromarine but I don't really believe he was ever top 10 worldwide in terms of skill. Even after his one Kato top 4 he didn't follow that up with any major results suggesting it was a fluke. Regardless even 2014 is late enough to show issues. Those first years were the most active years of SC2. Why would a system that by your own admission didn't work with 100,000 data points suddenly work with 200,000-300,000? Numbers are estimated based on aligulac's search function. The timeframe for Maru would be 2012-2017 and the metrics I already specified were efficiency, average placement, and percentage of won tournaments. The way you measure those Maru would have been better off being an unknown/inactive player who never qualified for anything during that era because then his 2018 forward timeframe would produce much better results in those metrics. Any measuring system that determines being inactive superior to Maru's 2012-2017 results is crazy. Prep format, offline in Korea, having to travel back and forth to events and simply a higher sample size. Some of Serral's impressive win rate years vs Koreans are based on a very small number of series. For example 2018 I doubt he could have kept that winrrate as high if he played more than 28 series. On the flipside, it’s not even a versus Korean win rate really, it’s almost a win rate specifically against top, top tier Koreans. I think there’s a case that the form he was in, 2018 Serral could have raised his win rate if he was unleashed on a bunch of perpetual GSL Ro32/Ro16 at best level Koreans In some subsequent years I think that’s less the case, in 2018 the opposition really was somewhat filtered that way. I’m not sure you take Serral’s opponents that year, have another Korean play those sets and have any of them get that win rate. He has quite a few years that are streets ahead of most by that specific metric, and he’s rarely playing the lower tier Koreans. What ifs are always tricky, but I'm not sold he'd increase his winrate if he had played in gsl that year. It's just likely he would've been given the reynor treatment and kicked out in groups. For all the talk about his good winrate against koreans, in 2018 his matches against koreans (in his best case scenario of no gsl level prep) were competitive but he wasn't dominating them. Iem 2018: went 3-2 against trap in the ro8 and 0-3 against classic in the semis Wesg 2018: went 0-3 against maru in the semis His best result by far that whole year was blizzcon, and even then stats took 2 games off of him.
GSL is tough, as Reynor has indeed found although I’ve long thought Serral’s just that little bit better at being consistent and hanging in there.
Regardless, whether Serral would/wouldn’t do well in GSL, my point was more he doesn’t tend to play the lower calibre of that cohort. Which Korean players do get to do more often. Not just in GSL itself but qualifiers, and qualifiers for international tournaments as well
I mean if we actually logged a versus foreigner metric and Maru’s basically only playing Serral, Clem, Reynor and a few others from the tier down, it’ll look quite a bit different than the range he gets now. And would be bigger still if he played those further down.
|
Interesting post.
Did not know Rain had that high winrate tbqh
|
Northern Ireland25035 Posts
On June 02 2025 12:09 goody153 wrote:
Interesting post.
Did not know Rain had that high winrate tbqh
Guy was pretty damn good, he’s gotta be up there as amongst the most talented overall RTS players considering his success across two quite different titles.
|
United States1847 Posts
On June 02 2025 12:30 WombaT wrote:Show nested quote +On June 02 2025 12:09 goody153 wrote:
Interesting post.
Did not know Rain had that high winrate tbqh Guy was pretty damn good, he’s gotta be up there as amongst the most talented overall RTS players considering his success across two quite different titles.
Of the players who won the most KIL from the period of time stretching from KeSPA switching to SC2 to the end of Hots (summer of 2012-2015), (Classic, Inno, Maru, Rain and Life all won two) only Inno and Rain also reached a third KIL final. Such a short career, but surely an excellent one.
|
Is being the greatest of all time in the worst expansion of all time with the worst balance council of all time and the least supported professional scene of all time and the most stagnant professional scene of all time even important?
I'm being facetious, I just don't think there's anything to gain from repeating this analysis ad nauseam...
|
|
|
|
|
|
EPT
