So this came up in the LR thread for today (whould've thunk it). Is Stephano Code S level or not? So I thought, well, with my statistical model I can maybe answer that question in an objective manner, so here's what I've done.
Code S currently contains 32 players, including people such as Huk who are generally agreed upon by many not to belong there. Nevertheless he is there. These 32 players will be our benchmark group.
Then we take a bunch of other players counted among the best in the world, let them play 32 hypothetical games against everyone in the benchmark group, and record their expected score. This is a number between 0 and 32.
For players already in the benchmark group, they get a "free" 0.5 because that's their expected score against themselves. Thus their score will be somewhere between 0.5 and 31.5.
Before I present the list I'd like to first point out that this rating considers games, not tournaments, and it weighs recent performance quite heavily compared to past ones. Thus, if your guy is a serial GSL code S winner from the bygone age, such as the holy trinity of MC, NesTea and Mvp, well that doesn't matter here.
Note also that this list is based on the top 150 rated players (active and inactive), and the code S players. The ranking can change a little bit in the lower levels if I expand this, because players who have high ratings against T and Z but low ratings against P will do better than normal, since the benchmark group contains so few Protosses.
If you have questions about the math you can read the FAQ or just browse the website where I play around with all this stuff.
So what do we find?
1. Life 22.669 (code S) 2. First 21.351
First may be the best player currently not in Code S.
This concludes the list of the players who can expect at least an even score in the code S with the current lineup. I have skipped Dark, who sneaks in on place 22 between San and Sen, but he is inactive after having played no rated games in eight weeks.
Perhaps surprising is the appearance of people like Maru, Vortix, San, Sen, JangBi and ForGG, who are not usually brought up in these discussions.
31. YoDa 15.768 (code S) 32. EffOrt 15.730
That is the size of the code S. Would you consider the above lineup an improvement to the current one? Probably not, I'm guessing. YuGiOh sneaks in on 31st place just ahead of YoDa, but like Dark, he is inactive (will return next list though, as he some games from January 31st).
33. Flash 15.506 34. RorO 15.499 (code S)
RorO is the strongest player currently in code S who has a "dubious" claim to his spot. Fun fact: CoCa's rating puts him around here, but he is of course also inactive.
So that concludes our top 50. "But wait!", you might say, "what in the name of Dustin Browder is Giantt doing there?". Yeah, good question, I don't know either. He is probably helped out by the fact that he has a very low ZvP rating, but there are so few Protosses in code S, so he gets a bit of an artificial boost.
"But wait!", you might say again, and I'll answer "what now?". "There are code S players missing!" Indeed there are. Some end up quite far down the list, it turns out.
In the end I must say it is extremely impressive for a tournament with a relatively rigid qualification structure to be able to gather 16 of the top 32, and 32 of the top ~130. It's much easier to assemble ridiculous lineups in invitationals.
Can you explain in detail how these scores were calculated? Just off TLPD ELO or something? I feel like Jaedong certainly does not belong in Code S yet, let alone top 8, and I'm shocked that his score would be so high.
On February 01 2013 21:43 randomsaint wrote: Can you explain in detail how these scores were calculated? Just off TLPD ELO or something? I feel like Jaedong certainly does not belong in Code S yet, let alone top 8, and I'm shocked that his score would be so high.
On February 01 2013 21:43 randomsaint wrote: Can you explain in detail how these scores were calculated? Just off TLPD ELO or something? I feel like Jaedong certainly does not belong in Code S yet, let alone top 8, and I'm shocked that his score would be so high.
Read the FAQ. If you have more questions after, just ask.
On February 01 2013 21:43 randomsaint wrote: Can you explain in detail how these scores were calculated? Just off TLPD ELO or something? I feel like Jaedong certainly does not belong in Code S yet, let alone top 8, and I'm shocked that his score would be so high.
Read the FAQ. If you have more questions after, just ask.
You should put that in the op (unless you did and I just don't see it), I also had no idea what it was until I saw this post.
On February 01 2013 21:43 randomsaint wrote: Can you explain in detail how these scores were calculated? Just off TLPD ELO or something? I feel like Jaedong certainly does not belong in Code S yet, let alone top 8, and I'm shocked that his score would be so high.
Read the FAQ. If you have more questions after, just ask.
You should put that in the op (unless you did and I just don't see it), I also had no idea what it was until I saw this post.
Sorry, I'll put it in there. I just assumed most people had heard about this by now.
Awesome research by the way, very level headed compared to a lot of threads that skew data to "prove" something is imba or whatever. While Code S is hard to qualify for, it doesn't make it foolproof. Look at Code A's YongHwa vs Life. If YongHwa had fought anyone else he would've made it through, instead he had to play up and downs while suffering with some kind of eye infection.And even the up and downs themselves can have groups of death as well as soft groups (Hence HuK making it through, no offense to him, he still played great that night). So basically, while GSL does what it can to get the cream of the crop, even it isn't immune to bracket luck.
These four are not in code S, of course, but I don't think many people would disagree that they show the level, gender be damned.
Eh, give players time to figure out how Scarlett is able to tech up so quickly and they'll come up with plenty of 2 or 3 base all ins to kill her. Same for JaeDong, as long as players mentally prepare themselves for treating his creep like lava then they can deny his biggest asset. But since we're using Code S level as a skill level instead of just how they would do in Code S, I'm in agreement
.... I swear we have a similar thread like this atleast once a week. People sure do love there statistics these days... Kind of interesting though I guess (surprised Lucky is on that list)
Speaking of recent LRs, a lot of people in the PL LR today were rehearsing the common line that EG-TL is losing not because of the quality of players but because they don't know how to play in PL format. What's interesting to me about this is that not only is there no evidence provided that this is the case, but they also don't even provide enough evidence to show that there is a phenomenon that needs explaining.
That is, I haven't seen it shown that the players on EG-TL have actually under-performed compared to what you would expect them to do given the matches each has played.
Have you checked out the data on this, TheBB? If not, I think it'd be a pretty interesting project.
Code S level has as much meaning as Grand Master level. Not really anything, because everyone has so different opinion of what it means, so there is not a majority opinion so basically it means nothing.
Impressed that you actually put this together after the discussion in today's GSL LR-thread. Great work and definitely interesting statistics. I am not sure about how exactly some of these results come to be but I'll have a look at the FAQ's.
would like Code A players listed in the spoiler to give an overview of GOM's overall pool of players since Code A will contain some Code S players or soon to be Code S players..
There are too many variables in the game that affects the outcome of the match overall. I don't think you can rank the players like this but you can definitely rank the top 5/10 or so.... with that said, I don't think First belongs in top 5, let alone top 20.... same with jaedong and scarlett.
On February 01 2013 22:25 goswser wrote: Giantt is at 41?
I addressed this question in the post. Since code S currently has so few Protosses, players who have strong vT and vZ matchups compared to their vP will benefit. Giantt is one of those. (You are another, as it turns out, but not nearly as strongly.)
On February 01 2013 22:35 foxj wrote: YugOh, Lyn aren't inactive since YugiOh has just joined Root ? And Lyn is playing for a Chinese team as well
Inactive doesn't mean what you think it means. FAQ.
If a player has not played any games for four periods (eight weeks, or about two months), he or she is removed from the list. The entry is still there, and will be taken into account if the player plays a new game some time in the future. It's only kept from the published list to keep it from filling up with uncertain and possibly irrelevant data.
The same happens if your rating uncertainty goes over 200.
If a player has not played any games for four periods (eight weeks, or about two months), he or she is removed from the list. The entry is still there, and will be taken into account if the player plays a new game some time in the future. It's only kept from the published list to keep it from filling up with uncertain and possibly irrelevant data.
The same happens if your rating uncertainty goes over 200.
Edit: God DAMMIT!
Haha okay I get it :D Sorry for not reading FAQ T~T
Eh, give players time to figure out how Scarlett is able to tech up so quickly and they'll come up with plenty of 2 or 3 base all ins to kill her. Same for JaeDong, as long as players mentally prepare themselves for treating his creep like lava then they can deny his biggest asset. But since we're using Code S level as a skill level instead of just how they would do in Code S, I'm in agreement
Kinda like what we saw from Stephano today; at least in the 2nd set, Innovation had to do relatively all-ins (proxy 2rax then next game 6rax timing) because if Stephano gets into the lategame, hes almost unstoppable utilizing his talent in strategic positioning and micro. Theres no better zerg in the world IMO in the lategame, but Stephano plays extra greedy early and Innovation took advantage of that. The beauty of this game, you need to adapt to your opponents weaknesses and exploit them, cant just play standard and win every single time with the most solid buildorder like Flash did for so many years. Flash isnt so dominant in SC2 and thats a big reason why IMO, you can say its "less-mechanically demanding" and obviously thats true but it results in the overall package deciding whos the best in the world.
On the Code S argument, I think theres a good 50-ish people who can be considered "Code S Quality" and its not only subject to the players currently in this season. Gotta look at who they consistantly win against.
Being consitent in CodeS means alot more than winning random tournaments like MLG becasue it's alot more demanding on you're actual play rather than enducance.
Prepairing for matches, players and being able to adapt and play on the fly is what makes truly great players (MVP for example) and it's that kind of play that's most rewarded in CodeS
Eh, give players time to figure out how Scarlett is able to tech up so quickly and they'll come up with plenty of 2 or 3 base all ins to kill her. Same for JaeDong, as long as players mentally prepare themselves for treating his creep like lava then they can deny his biggest asset. But since we're using Code S level as a skill level instead of just how they would do in Code S, I'm in agreement
Kinda like what we saw from Stephano today; at least in the 2nd set, Innovation had to do relatively all-ins (proxy 2rax then next game 6rax timing) because if Stephano gets into the lategame, hes almost unstoppable utilizing his talent in strategic positioning and micro. Theres no better zerg in the world IMO in the lategame, but Stephano plays extra greedy early and Innovation took advantage of that. The beauty of this game, you need to adapt to your opponents weaknesses and exploit them, cant just play standard and win every single time with the most solid buildorder like Flash did for so many years. Flash isnt so dominant in SC2 and thats a big reason why IMO, you can say its "less-mechanically demanding" and obviously thats true but it results in the overall package deciding whos the best in the world.
On the Code S argument, I think theres a good 50-ish people who can be considered "Code S Quality" and its not only subject to the players currently in this season. Gotta look at who they consistantly win against.
Well, I think the weakness of being extra greedy is related to the late game dominance. Pro players don't scout for a reason. They want to eek out that Eco advantage. Of course, it backfires if your opponent is aggressive and you get punished. Not scouting with a drone means more minerals earlier which means your can drone and tech early. Of course, we are talking about seconds here. But at the pro levels, 'stealing' a few seconds here and there means so much. Having an upgrade timing can be the different between an even trade to one where you gain 10 supply. Same with getting just a couple units more out for an engagement.
Having 5-6 more army supply at an engagement could mean you get another 6 supply advantage. And with your 10 supply advantage, it means your next engagement will get another 10 supply and it just snowballs.
So while players like MKP and Stephano SHOULD play safer. We have to keep in mind that it will make a difference in their late game. 1 drone at the 90 second mark is very different from 1 drone at the 10 minute mark.
The problem with this "ranking" is that some players play against really really high level players all the time, while some don't and others don't really ever play against the top, but are still higher.
This does a lot to highlight some of the issues sc2 is facing right now. The way the races are designed is a huge issue and it really shows in these statistics. Zerg really only players the game one way. The way most Zerg games play out is so predictable that it is very easy for Zergs to focus their practice. Protoss players are similar in that they rely the most on the current metagame to win, so they tend to play very close to each other. Protoss players do have more potential strats to choose from though. Terran players really dont have a catch all style and instead rely heavily on extremely well practiced build orders and timings to exploit their opponents, something Zerg can't do and Protoss can only do if the current meta conditions are right for it.
This leads to a lot of the extremely skilled Terran to be able to compete in tourneys like Code S/Code A where you have time to prepare for each opponent. Terran benefits the most from being able to carefully devise a plan to exploit something their opponent might be doing either intentionally or unintentionally. The reason the list is so top heavy with Zergs is because in tourneys without long gaps between games Zergs are at an advantage as everything will be played "as is" with the games being a lot more generic in terms of potential strats.
This table does little to show imbalance or even skill, but it does expose some of the core design flaws sc2 is currently showing (and HoTS doesn't seem to offer much in the way of improving that)
Sorry, maybe it's just me, or you put Scarlet over Leenock. Huh? I'm not seeing things? Oh, wait, there's Taeja, Polt and Curious standing over Vortix.
Players win over people that are 300 positions behind them and lose to ones that are 10-20 positions behind them, yet still their rating grows. To put it simply, you're a GM winning against mid-masters and losing to fellow GM, yet your rating grows.
On February 01 2013 23:51 HumpingHydra wrote: My question is has anyone used this to determine their voting pattern in Liquibets? What would the result be?
TheBB should write a bot that uses aligulac for liquibets and fantasy proleague. :D
On February 02 2013 00:18 Hiea wrote: The problem with this "ranking" is that some players play against really really high level players all the time, while some don't and others don't really ever play against the top, but are still higher.
Yes i agree. Taking the current 32 Code S Paricipants as a benchmark is pretty random imo and leads to some strange rankings like Scarlett far above Taeja for example. The benchmark should be something like what is a average win/loss rate in combination with the amount of appearences in Code S. Because being Code S-Level means to be able to repeat success in this league over a certain period of time.
On February 01 2013 23:51 HumpingHydra wrote: My question is has anyone used this to determine their voting pattern in Liquibets? What would the result be?
TheBB should write a bot that uses aligulac for liquibets and fantasy proleague. :D
I've considered it. Well, not a bot, but just doing it manually.
My gut instinct is that a good liquibetter would probably do better, but that the ratings are for sure more accurate than the people who write the TL predictions.
On February 02 2013 00:18 Hiea wrote: The problem with this "ranking" is that some players play against really really high level players all the time, while some don't and others don't really ever play against the top, but are still higher.
Yes i agree. Taking the current 32 Code S Paricipants as a benchmark is pretty random imo and leads to some strange rankings like Scarlett far above Taeja for example. The benchmark should be something like what is a average win/loss rate in combination with the amount of appearences in Code S. Because being Code S-Level means to be able to repeat success in this league over a certain period of time.
Not what I meant, it's more that someone like Giant, has he ever competed against a korean? I think he took a game off Nerchio, but besides that...
Why does he rank higher than someone like Mvp? It shows that there is gigantic flaws in this list, and if I ever see it used as fact somewhere I will die a little inside.
And when you think about how a lot of the KeSPA players don't play the top, top tier, they play eachother and that means a lot of B teams, boosting their ranks, and koreans who travel to foreign tournaments a lot will also rise higher than the koreans who compete almost only in GSL.
In my opinion, you have code s level when you're able to stay in the league over a longer period of time AND/OR peaking in a single season. (like jiakji and seed, who aren't on code s level atm) So basically almost every player in code S actually belongs there in some way, people like Huk and Hack being examples for exceptions.
But the whole talk about "levels" is nonsensical. The best players are staying on top. Just take DRG as an example, even though he might have had his off-seasons, he has remained a top player since 2011.
You also just can't neglect the fact that the level of competition in Korea is insane. And the current format is pretty brutal when you're in a mini slump. (just look at Rain)
On February 02 2013 00:30 DidYuhim wrote: Sorry, maybe it's just me, or you put Scarlet over Leenock. Huh? I'm not seeing things? Oh, wait, there's Taeja, Polt and Curious standing over Vortix.
Players win over people that are 300 positions behind them and lose to ones that are 10-20 positions behind them, yet still their rating grows. To put it simply, you're a GM winning against mid-masters and losing to fellow GM, yet your rating grows.
It's a model, it's not who chose to put who above who, it's what this particular model think the number shows. SC2 model seriously suffer from SSS problem as there needs to have dozens if not hundreds matches for each player for model be considerably accurate.
On February 02 2013 00:18 Hiea wrote: The problem with this "ranking" is that some players play against really really high level players all the time, while some don't and others don't really ever play against the top, but are still higher.
Yes i agree. Taking the current 32 Code S Paricipants as a benchmark is pretty random imo and leads to some strange rankings like Scarlett far above Taeja for example. The benchmark should be something like what is a average win/loss rate in combination with the amount of appearences in Code S. Because being Code S-Level means to be able to repeat success in this league over a certain period of time.
Taeja didnt do very well in the past few month (injury). In the OP, it is said the statistics are more influenced by recent results than by old ones. I guess the Taeja dominance of this summer does not matter anymore in those stats.
These four are not in code S, of course, but I don't think many people would disagree that they show the level, gender be damned.
I'm pretty confident in saying that viOLet is the only Code S worthy player out of these 4.
Monster once was, in a time long ago which has now gone to myth rumours say thorzain belonged to a once legendary team known as "mousesports" back when monster was a legit zerg.
On February 02 2013 00:18 Hiea wrote: The problem with this "ranking" is that some players play against really really high level players all the time, while some don't and others don't really ever play against the top, but are still higher.
Yes i agree. Taking the current 32 Code S Paricipants as a benchmark is pretty random imo and leads to some strange rankings like Scarlett far above Taeja for example. The benchmark should be something like what is a average win/loss rate in combination with the amount of appearences in Code S. Because being Code S-Level means to be able to repeat success in this league over a certain period of time.
Not what I meant, it's more that someone like Giant, has he ever competed against a korean? I think he took a game off Nerchio, but besides that...
Why does he rank higher than someone like Mvp? It shows that there is gigantic flaws in this list, and if I ever see it used as fact somewhere I will die a little inside.
It is what it is. It's a fact that this is what you get when you run the numbers the way I did. Anything outside of that is extrapolation, which may or may not be of much use. Anyone should be careful when they use statistics, and this is no exception.
But please, has Giantt ever competed against a Korean? The list is on the fucking website. It's not rocket science to actually go there and check. http://aligulac.com/players/294/results/
He has an even score against Hyun and 2-0 against TheStC for example.
And can I just say, why would that be the defining characteristic? A passport doesn't make you good at StarCraft. You should be asking about his results against highly rated players, not against Koreans.
He does NOT rank higher than Mvp. See here. They have almost the same rating (Mvp slightly higher). Like I have explanied, Giantt ranks higher than Mvp on the list given in the OP because it uses a different ranking system (performance against a pool of mostly Zergs and Terrans, matchups in which Giantt plays well).
But still, Mvp is almost always pointed out as a flaw in the system and I'm getting kinda tired of it. He hardly plays any more and his recent results are middling at best. http://aligulac.com/players/13/results/
He's just not very good and that's why he's not rated highly. Get over it.
Of course, as "code S level" goes, that can mean anything. The OP takes it to mean "would probably do ok in this season of code S", but that shows only one side of the story.
On February 02 2013 00:23 Filter wrote: This does a lot to highlight some of the issues sc2 is facing right now. The way the races are designed is a huge issue and it really shows in these statistics. Zerg really only players the game one way. The way most Zerg games play out is so predictable that it is very easy for Zergs to focus their practice. Protoss players are similar in that they rely the most on the current metagame to win, so they tend to play very close to each other. Protoss players do have more potential strats to choose from though. Terran players really dont have a catch all style and instead rely heavily on extremely well practiced build orders and timings to exploit their opponents, something Zerg can't do and Protoss can only do if the current meta conditions are right for it.
This leads to a lot of the extremely skilled Terran to be able to compete in tourneys like Code S/Code A where you have time to prepare for each opponent. Terran benefits the most from being able to carefully devise a plan to exploit something their opponent might be doing either intentionally or unintentionally. The reason the list is so top heavy with Zergs is because in tourneys without long gaps between games Zergs are at an advantage as everything will be played "as is" with the games being a lot more generic in terms of potential strats.
This table does little to show imbalance or even skill, but it does expose some of the core design flaws sc2 is currently showing (and HoTS doesn't seem to offer much in the way of improving that)
Why do you think this is a design flaw? I think it's fine if every race has his own style, Z being more predictable and passive-macro oriented, Protoss trying to abuse metagame as much as possible and Terran being more of a timing oriented race.
On February 02 2013 00:23 Filter wrote: This does a lot to highlight some of the issues sc2 is facing right now. The way the races are designed is a huge issue and it really shows in these statistics. Zerg really only players the game one way. The way most Zerg games play out is so predictable that it is very easy for Zergs to focus their practice. Protoss players are similar in that they rely the most on the current metagame to win, so they tend to play very close to each other. Protoss players do have more potential strats to choose from though. Terran players really dont have a catch all style and instead rely heavily on extremely well practiced build orders and timings to exploit their opponents, something Zerg can't do and Protoss can only do if the current meta conditions are right for it.
This leads to a lot of the extremely skilled Terran to be able to compete in tourneys like Code S/Code A where you have time to prepare for each opponent. Terran benefits the most from being able to carefully devise a plan to exploit something their opponent might be doing either intentionally or unintentionally. The reason the list is so top heavy with Zergs is because in tourneys without long gaps between games Zergs are at an advantage as everything will be played "as is" with the games being a lot more generic in terms of potential strats.
This table does little to show imbalance or even skill, but it does expose some of the core design flaws sc2 is currently showing (and HoTS doesn't seem to offer much in the way of improving that)
A really interesting theory, and in my opinion a good writeup of how the game plays out.
That Protoss are so relient on metagame trends is something I do dislike. I much prefer Terran's much more varied approach to the game. To me, it looks like this is something I will get more of for Protoss in HotS, where Protoss air is given some more/better options.
I think there are far too many 'Code S level' players out there. In Tennis, out of the last 40 premier tournaments, there are only five players that have won. The second place finishes are mostly these same players also. The number of players that are legitimately competitive in any given era is about a dozen. In Starcraft II it's rare to repeat a victory, it's even not unusual for the Ro8 to completely differ from tournament to tournament. I think there are about a hundred players that all can take Bo3's off each other. If you look at the full list in the OP, there are just so many good players out there, it's kind of insane.
I won't say it's always like this: there have been times where especially high level zergs were really stable and dominant, for instance DRG, Symbol, Leenock and Life have all had runs where they seemed unstoppable. MVP, MMA, Taeja have had similar runs. And that's a good thing to have for the game, but I don't think it happens often enough, and those players are still vulnerable. I don't think we'll ever see the same amount of consistency that Flash and Jaedong could achieve in Brood War. (unless Blizzard changes the game) I think in the future we will continue to see one-hit wonders.
On February 01 2013 23:51 HumpingHydra wrote: My question is has anyone used this to determine their voting pattern in Liquibets? What would the result be?
TheBB should write a bot that uses aligulac for liquibets and fantasy proleague. :D
I've considered it. Well, not a bot, but just doing it manually.
My gut instinct is that a good liquibetter would probably do better, but that the ratings are for sure more accurate than the people who write the TL predictions.
Not that hard since the TL predictions writers are almost as biased as Tastosis .
"Code S level" means you're in Code S. It includes the people who lose in ro32 but get back to Code S.
I think the difficult concept to grasp is that it is a fluid and dynamic label. Jjakji is not Code S level. He WAS, and he won the GSL, but he isn't right now. On the other side, Losira is Code S level. He wasn't for a long time, but he's back.
Stephano is in a strange position because he didn't "earn" his Code S spot, it was given to him as a seed. So the only test of whether he is a Code S level player is if he can get back in. If he can't, then he's not Code S level. You can make up excuses for him, but the same standard gets applied to everyone else, even beloved players like MVP and Leenock.
On February 02 2013 01:17 Grumbels wrote: I think there are far too many 'Code S level' players out there. In Tennis, out of the last 40 premier tournaments, there are only five players that have won. The second place finishes are mostly these same players also. The number of players that are legitimately competitive in any given era is about a dozen. In Starcraft II it's rare to repeat a victory, it's even not unusual for the Ro8 to completely differ from tournament to tournament. I think there are about a hundred players that all can take Bo3's off each other. If you look at the full list in the OP, there are just so many good players out there, it's kind of insane.
I won't say it's always like this: there have been times where especially high level zergs were really stable and dominant, for instance DRG, Symbol, Leenock and Life have all had runs where they seemed unstoppable. MVP, MMA, Taeja have had similar runs. And that's a good thing to have for the game, but I don't think it happens often enough, and those players are still vulnerable. I don't think we'll ever see the same amount of consistency that Flash and Jaedong could achieve in Brood War. (unless Blizzard changes the game) I think in the future we will continue to see one-hit wonders.
You forgot wc3, ~80% of all premier tournaments were won by Grubby, Moon, ToD, Lyn or Sky since 2004 until sc2 release. I really miss this consistency in sc2. sc2:wol is really super random..
On February 02 2013 01:25 coverpunch wrote: "Code S level" means you're in Code S. It includes the people who lose in ro32 but get back to Code S.
On February 02 2013 01:33 AimlessAmoeba wrote: I think "Code S level" doesn't mean anything at all - You're either Code S, or you're not. It really is that simple.
Code S currently contains 32 players, including people such as Huk who are generally agreed upon by many not to belong there. Nevertheless he is there. These 32 players will be our benchmark group.
As far as I'm concerned, if you're in Code S, you're definitely a Code S player. I guess I pretty much go by the definition in that regard. You've earned the title, since you're in Code S. The only exceptions perhaps would be if you received a free seed into Code S and then dropped out rather quickly, showing that perhaps you didn't deserve the spot in the first place.
Now, certainly many Code A players have the potential to be in Code S, and perhaps the only thing holding them back was the fact that they had a tough bracket or bad draw or some unfortunate luck. And that's where it gets a bit subjective and gray for me, but I feel that within two or three seasons, they'll either get to Code S anyway (thus proving they deserve it) or they'll remain in Code A/ drop out, (thus proving they don't deserve it).
I think it's important to reserve judgment for some players. Some who are on the cusp don't need to be "clearly" declared as Code S (or anything) just yet.
On February 02 2013 00:18 Hiea wrote: The problem with this "ranking" is that some players play against really really high level players all the time, while some don't and others don't really ever play against the top, but are still higher.
Yes i agree. Taking the current 32 Code S Paricipants as a benchmark is pretty random imo and leads to some strange rankings like Scarlett far above Taeja for example. The benchmark should be something like what is a average win/loss rate in combination with the amount of appearences in Code S. Because being Code S-Level means to be able to repeat success in this league over a certain period of time.
Taeja didnt do very well in the past few month (injury). In the OP, it is said the statistics are more influenced by recent results than by old ones. I guess the Taeja dominance of this summer does not matter anymore in those stats.
It was just an example... Taeja is participating Code S since s2 2012 and he is still in. He advanced from the first group stage in every season except season 5. He is clearly a Code S player. What did Scarlett (or other players) in GSL? The lack of longterm stats in this project makes its validity uncertain... in a year it will be better i think, but at the moment you cant determine if a player is code s level or not.
Another awesome post from our swiss, math-savvy nostradamus. Give this man a star already. :D
In my opinion, there are 3 possible "labels" you can give a player based on Code S:
(i) the first one is a "Code S player", i.e., one that currently competes in Code S, regardless of if he "deserves it". HuK, for instance, was a Code S player a couple of weeks ago, even though some question whether his skill is actually up to par.
(ii) the second one is a "Code S level player", which, to me, is a generic way of saying that a player has all the capacities (mainly mechanical and mental) to be paired against the best (korean) SC2 players in the world. This does not mean, however, that a player must advance from the ro48 or win in the U&D to prove his skill, especially if we consider that SC2 is a highly volatile game that harshly punishes the smallest of mistakes or build order choices.
(iii) the last one is a "Code S title contender", which is what a lot of people consider to be "Code S level". Having a chance to win it all takes a lot of korean-style preparation (oft overlooked by foreigners who participate in the GSL), but, while a very useful skill to have and master, I don't think it actually factors in how solid a player's overall skill level is. So, if a player isn't determined enough to put in the hours and prepare for his matches, it's safe to say he'll never be a title contender, but that doesn't mean his isn't "Code S material", to put it that way.
Pity it's mostly interpreted as anyone who is in code S, has been in code S recently, and anyone who has looked competent against the above, unless you're not Korean, then it doesn't count due to lack of opponent preparation
On February 02 2013 00:23 Filter wrote: This does a lot to highlight some of the issues sc2 is facing right now. The way the races are designed is a huge issue and it really shows in these statistics. Zerg really only players the game one way. The way most Zerg games play out is so predictable that it is very easy for Zergs to focus their practice. Protoss players are similar in that they rely the most on the current metagame to win, so they tend to play very close to each other. Protoss players do have more potential strats to choose from though. Terran players really dont have a catch all style and instead rely heavily on extremely well practiced build orders and timings to exploit their opponents, something Zerg can't do and Protoss can only do if the current meta conditions are right for it.
This leads to a lot of the extremely skilled Terran to be able to compete in tourneys like Code S/Code A where you have time to prepare for each opponent. Terran benefits the most from being able to carefully devise a plan to exploit something their opponent might be doing either intentionally or unintentionally. The reason the list is so top heavy with Zergs is because in tourneys without long gaps between games Zergs are at an advantage as everything will be played "as is" with the games being a lot more generic in terms of potential strats.
This table does little to show imbalance or even skill, but it does expose some of the core design flaws sc2 is currently showing (and HoTS doesn't seem to offer much in the way of improving that)
Why do you think this is a design flaw? I think it's fine if every race has his own style, Z being more predictable and passive-macro oriented, Protoss trying to abuse metagame as much as possible and Terran being more of a timing oriented race.
Races having their own styles is great for the game, the problem is that only one race is really capable of showing any real character and flair for players. Even then Terran is being pushed away from that are the game gets more and more figured out. MKP, MVP and MMA used to all have very distinct and different styles. This creates fans and makes games a lot more interesting. All Zergs look exactly the same when they play, Protoss players all look the same with some variations in execution being evident but nothing stylistic showing up anymore. This leads to a very stale games to watch, and even more stale games to play.
When you look past the pro game at the ladder games that you and I probably play, this starts to show up more and more. I remember being absolutely inspired watching MMA play and trying to integrate his builds and moves into my own play on ladder, and having a blast failing and succeeding. Zerg and Protoss players don't really have hero's to emulate and create passion. You could pick any of the top ten Zergs in the world and try to emulate their styles and end up at exactly the same point.
The difference is not in racial mechanics, it's simply a huge design problem and something that has not been addressed at all in HoTS. Terran builds/styles are based on extremely careful resource allocation and smoothly transitioning from one unit to the next. The most extreme example is going with a 1/1/1 banshee expand into a quick medivac drop with very early stim. Every building chains to the next and each unit has a very specific purpose on the field. Miss something by 30 seconds and the entire build is basically thrown out. The build has a very clear plan from start to finish and everything links together in a very clean manner. You dont have the ability to suddenly go cloaked banshee, or go for a heavy mech army because you won't have the gas to support it. Your entire infrastructure is based on a bio transition after the initial banshee and practice/execution is vital to having success with this type of build.
I'm okay with differences in how the races handle, I don't expect Zerg to be making addons to roach warrens and producing roaches out of them. At the same time though Zerg units do not link together at all, and this is a huge problem. Nobody uses hydra's because they're expensive and only situationally useful. If hydra's were capable of being morphed into something else once you researched the upgrade, they would see a lot more play. Zerg can quite literally shift through the units with no thought up front and simply make whatever suits them at the time. This leads to almost no reactive play from a Zerg player, and lets them dictate how the game will be played to his opponents. That in turn leads to stale "best" strategies that are used by all Zergs in pro play and on the ladder. Even things that you would think would have synergy and lead into each other do not. Zerglings and Zergling upgrades don't nudge a player in the direction of ultra's, in fact Zerg melee upgrades are more effective on broodlords than they are on ultra's so most people go that route. With no linking of techs Zerg players are free to just make the best units available instead of having to choose between something they have the research/units for already and something that might be better for their current situation but that they don't have the upgrades to use to it's full potential. Zerglings and their upgrades are effective for both T3 units. Roaches don't need +attack to be effective and hydra's get railed by everything that's in the game to kill marines + they have no followup synergy to help them out. This means that a zerg can quite comfortably make whatever units he wants/needs no matter how he has chosen to handle his upgrades.
Protoss players are forced to play the metagame, or lose. If you go for an oldschool 3rax (2tech,1reactor) against toss players on ladder you'll win 80% of your games because in order to keep up with what most Terran players do they have to 1gate expand or go Nexus first. Do anything but forge expand against Zerg and you lose. Things are more interesting in PvP but at the same time that matchup is basically rock paper scissors followed by war of the worlds. If a Protoss player doesn't play what is the current meta in PvT and PvZ and they face a standard player they lose. If they face somebody who players with older meta's and they player current they lose. Ever wonder why I haven't gone near Toss for some tutorial videos? that's why. I wouldn't go near the race if a I was a competitive player because the amount of control I would actually have over winning and losing would basically be zero. MC is the perfect example of this, for the last two years he's changed his style 3 or 4 times per year. He wins a tourney then he loses one in the first round with such consistency it's amazing. MC has also changed his style over and over because he simply has to to stay competitive. Some players come and have huge showings then drop off the map to never be seen again and the biggest reason is they bring some awesome new style that nobody can deal with, then it gets figured out and they fail to adapt to the new meta.
On February 02 2013 02:10 Figgy wrote: Also, based off your points system only the top 16 should only have even a tiny bit of a chance of being above a 16 rating (because that's half the games played vs Code S players and the only amount possible to make it out), and no one below the top 32 should have above a .5 rating (1 in 32 chance of winning a game). I seriously don't even know how you came up with this BS because the statistics are completely whack. Your current statistics is comparing all 150 people against all 150 people, instead of the top 32 against the rest. Honestly, who cares how many people crush TLO or Johnnyrecco. We are talking about Code S
You can't possibly have read anything I wrote. The numbers compare all 150 people against the current code S players. Not the top 32 and not the full 150. This is spelled out clearly in the OP. Take the average of the code S players, you should get 16 exactly.
Some of your other concerns are legitimate and some are ridiculous. That's fine, but you can take your attitude and stuff it.
Wow I always knew this guy's rating system was bad, but this just proves it. No matter how defensive he gets defending his stats cause well he designed it, doesn't change the fact that his system is bad and you can't rank players by a statistical system that easy in sc2.
On February 02 2013 02:55 Canucklehead wrote: Wow I always knew this guy's rating system was bad, but this just proves it. No matter how defensive he gets defending his stats cause well he designed it, doesn't change the fact that his system is bad and you can't rank players by a statistical system that easy in sc2.
I think the idea is neat on paper (having a system that can help predict results) but in practice it isn't as efficient. For example, Innovation was predicate to have a 75% chance to get 3rd or 4th, basically not getting out of the group. Yet he managed to get second.
The system of course doesn't take into account stuff such as amount of practice / how they practice (stephano said he just massed ladder games to train for GSL, compare and contrast that with the GSL method of studying an opponent's playstyle / ect).
Love the work you've put in, but tbh "Code S" haven't been "Code S" since the bw days. I mean c'mon, one season they're "Code S champion", the next season they go out in ro32.
On February 02 2013 01:33 AimlessAmoeba wrote: I think "Code S level" doesn't mean anything at all - You're either Code S, or you're not. It really is that simple.
It was a rhetorical question.
Just so that's clear.
On February 02 2013 01:48 TheBB wrote: Is Stephano Code S level or not? So I thought, well, with my statistical model I can maybe answer that question in an objective manner, so here's what I've done.
On February 02 2013 02:10 Figgy wrote: These stats are worthless.
You need to compare players games vs other Code S players ONLY for these to make any difference.
First beating on some random foreign scrubs does not mean he'll win 22 out of 31 games against every Code S player in a row. It means he'll win 22 out of 31 games against random foreign scrubs with a couple high level koreans in the middle.
Foreign results are part of why these ratings are currently a joke, because foreigners get such high elo beating unknown foreigners due to lack of games.
Also, lulz on counting games that are over 1 year old. Like any of those matter.
Anyone seriously think Johnyrecco is going to win 1/3rd of his Code S games? Of course fucking not, he is going to go 0-4 and get laughed out of korea. Just like these stats. Has he even played a single Code S player in his career??? How can you possibly say he'd ever even take a game off a Korean let alone 1/3rd of his games in Code S.
TLO is an even bigger joke, according to these Statistics he would make it out of his group with a 50/50 chance.
You NEED to adjust your stats to only include people who have at least 15 games against current Code S players in the last 6 months. Otherwise these stats are trash.
Also, based off your points system only the top 16 should only have even a tiny bit of a chance of being above a 16 rating (because that's half the games played vs Code S players and the only amount possible to make it out), and no one below the top 32 should have above a .5 rating (1 in 32 chance of winning a game). I seriously don't even know how you came up with this BS because the statistics are completely whack. Your current statistics is comparing all 150 people against all 150 people, instead of the top 32 against the rest. Honestly, who cares how many people crush TLO or Johnnyrecco. We are talking about Code S
Maybe, just maybe, you should read up a little on what TheBB did here before you trash his work?
On February 02 2013 01:25 coverpunch wrote: "Code S level" means you're in Code S. It includes the people who lose in ro32 but get back to Code S.
On February 02 2013 01:33 AimlessAmoeba wrote: I think "Code S level" doesn't mean anything at all - You're either Code S, or you're not. It really is that simple.
On February 02 2013 01:48 TheBB wrote: Is Stephano Code S level or not? So I thought, well, with my statistical model I can maybe answer that question in an objective manner, so here's what I've done.
???
Rhetorical question = you ask a question but do not expect an answer, because you go on to either answer the question your self or you imply that the answer is so obvious that the question does not need answering. It was clearly a rhetorical question. The definition you gave in your answer is of course also a valid one, but with the statistical data he gathered, he just tried to give another viewpoint on this I found his work very interesting!
With every list TheBB posts, I am more and more convinced that ELO (also Glicko) is simply the wrong system for SC2! It was invented in the sixties for chess and that shows.
On February 02 2013 03:20 BadAim wrote: Love the work you've put in, but tbh "Code S" haven't been "Code S" since the bw days. I mean c'mon, one season they're "Code S champion", the next season they go out in ro32.
Code S is just a name that GOM created for the best SC2 players. I personally don't like it because it conflicts with the BW classes which better describe a player type.
B Class or B-Teamers - Practice Partners A Class - A-Teamers - A teams best starting line up players S Class - S-Class - The absolute best. S-Class players make A-Teamers look like B-Teamers
I believe the whole Code system is messed up and the class system is a better way of determining a players skill level. Simply because in order to be a "Code" you would have had to play in the GSL. You can be beating Code S players left and right outside of GSL and still not be labeled Code S just because you haven't made an appearance in the GSL. So, it's really conflicting and I wish we'd stop referring to the Code System when determining a players skill level. Because a player can be in Code S one season and back in Code A the next to being Code B the next. But still own Code S players outside the GSL.
Code S to me is an arbitrary measure of how well a player performs in GSL, simple as that. There are players who have plenty of online potential and have done well in foreign tournaments, but would you really consider ForGG, Jaedong, Rain or Maru competitive at a Code S level currently? My answer is no, but they have proven themselves competitive in online tournaments (like the aformentioned Giantt) or in foreign tournaments, which may lack the quantity of Code S players.
But in my opinion, it is a bit wrong to put every tournament- online and off-line, on the same level because there are so many factors that can change the results. Plus, because SC2 is a more volatile game than BW, it has a lot less of the consistency and skill that marked BW differently from SC2. Now comparing these two is a different topic all together, but it is important to note that Code S is such an unstable title because Code S Koreans get upset or defeated constantly, when in BW they would absolutely dominate any foreigner.
On February 02 2013 04:47 BlazeFury01 wrote: Code S is just a name that GOM created for the best SC2 players. I personally don't like it because it conflicts with the BW classes which better describe a player type.
B Class or B-Teamers - Practice Partners A Class - A-Teamers - A teams best starting line up players S Class - S-Class - The absolute best. S-Class players make A-Teamers look like B-Teamers
I believe the whole Code system is messed up and the class system is a better way of determining a players skill level. Simply because in order to be a "Code" you would have had to play in the GSL. You can be beating Code S players left and right outside of GSL and still not be labeled Code S just because you haven't made it there in the GSL. So, it's really conflicting and I wish we'd stop referring to the Code System.
i agree, gom depreciated the value of the word there was a time when "s" was associated with only a handful of most elite players...
"S Class - S-Class - The absolute best. S-Class players make A-Teamers look like B-Teamers" sadly that will never happen with sc2
SC2 has a large random factor. Any SC2 rating system is going to have its anomalies because of that. On top of that, Proleague (and GSL, too) is way more about preparing for one (or a few) specific opponents rather than being generally good at the game.
Given those caveats, I'd say the ranking given here is surprisingly accurate.
I don't want to discourage you from working on this rating system but this system is just missing too much information and factors to be worth anything at this particular moment.
As for defining what code S means, its quite simple, if you qualify for Code A without a seed (the old fashioned way), then your'e unquestionably code S level because you've essentially grinded your way through a pool that includes everyone else trying to do the same.
If you got seeded in, then it gets kind of tricky to say.
The rating is a reflection of recent past results and the OP's ranking added the flavor of expected results against the ro32 Code S players.
Does it have predictive value? Not really.
But the argument is that Stephano "deserves" to be in Code S because he isn't wildly out of line with other people by his ranking. And in fact HuK may also deserve Code S because he isn't even the lowest ranked player of the bunch. That would be Byun.
On February 02 2013 05:14 FuzzyJAM wrote: Without wishing to be a dick, this list is the best illustration that your stats are not close to being a good indicator of overall skill.
This.. A guy who's only ever played in online qualifiers/cups and even in those he only has about a 50% win-rate being in the top 40 players in the world is hilarious. I don't know how hard it would be but I think your rankings need to take into account the tournament they beat the Code S players at and give them more or less points depending on where it's at. Something like GSL > IPL > MLG > Dreamhack > all other lans > Big online tournaments > weekly cups/online qualifiers. Also maybe making it so they have to have at least 15-20 games played against the current Code S
On February 01 2013 23:51 HumpingHydra wrote: My question is has anyone used this to determine their voting pattern in Liquibets? What would the result be?
TheBB should write a bot that uses aligulac for liquibets and fantasy proleague. :D
I've considered it. Well, not a bot, but just doing it manually.
My gut instinct is that a good liquibetter would probably do better, but that the ratings are for sure more accurate than the people who write the TL predictions.
well, if you think about, even if everyone were complete noobs and picked bets at random, with enough of them you'd expect to be outbet by a large number of people. those people would then be labeled "good liquibetters" because they did better than you.
On February 02 2013 00:18 Hiea wrote: The problem with this "ranking" is that some players play against really really high level players all the time, while some don't and others don't really ever play against the top, but are still higher.
Yes i agree. Taking the current 32 Code S Paricipants as a benchmark is pretty random imo and leads to some strange rankings like Scarlett far above Taeja for example. The benchmark should be something like what is a average win/loss rate in combination with the amount of appearences in Code S. Because being Code S-Level means to be able to repeat success in this league over a certain period of time.
Not what I meant, it's more that someone like Giant, has he ever competed against a korean? I think he took a game off Nerchio, but besides that...
Why does he rank higher than someone like Mvp? It shows that there is gigantic flaws in this list, and if I ever see it used as fact somewhere I will die a little inside.
It is what it is. It's a fact that this is what you get when you run the numbers the way I did. Anything outside of that is extrapolation, which may or may not be of much use. Anyone should be careful when they use statistics, and this is no exception.
But please, has Giantt ever competed against a Korean? The list is on the fucking website. It's not rocket science to actually go there and check. http://aligulac.com/players/294/results/
He has an even score against Hyun and 2-0 against TheStC for example.
And can I just say, why would that be the defining characteristic? A passport doesn't make you good at StarCraft. You should be asking about his results against highly rated players, not against Koreans.
He does NOT rank higher than Mvp. See here. They have almost the same rating (Mvp slightly higher). Like I have explanied, Giantt ranks higher than Mvp on the list given in the OP because it uses a different ranking system (performance against a pool of mostly Zergs and Terrans, matchups in which Giantt plays well).
But still, Mvp is almost always pointed out as a flaw in the system and I'm getting kinda tired of it. He hardly plays any more and his recent results are middling at best. http://aligulac.com/players/13/results/
He's just not very good and that's why he's not rated highly. Get over it.
Of course, as "code S level" goes, that can mean anything. The OP takes it to mean "would probably do ok in this season of code S", but that shows only one side of the story.
The real question people is asking, is why was Giantt adjusted to 2301 rating vs Terran dispite never facing someone above 2k?
He got a massive artificial boost without facing any Terrans even close to that rating (And only beating one, TheSTC 2-0 who wasn't even 2k rated)
Why was First artifically boosted FIVE HUNDRED rating in all match ups dispite facing no one even close to that skill?
http://aligulac.com/players/39/period/76/ How does this extrapolation make any sense in any world? You boosted him to the 2nd best player on the planet dispite him not even meeting anyone in the top 25.
In ELO you cannot beat on someone who is lower than 400 points below you for a REASON. Because utterly destroying a random master leaguer 100 times in a row does not make you the best in the world regardless of your expected results against them.
On February 02 2013 05:14 FuzzyJAM wrote: Without wishing to be a dick, this list is the best illustration that your stats are not close to being a good indicator of overall skill.
There isn't really a skill list achievable through statistics out there that people would agree with anyway.
I mean on TLPD Puzzle, who hasn't played for months, is still higher than Soulkey. And tear, who just made Code A, is higher than many Code S players.
On February 02 2013 05:14 FuzzyJAM wrote: Without wishing to be a dick, this list is the best illustration that your stats are not close to being a good indicator of overall skill.
There isn't really a skill list achievable through statistics out there that people would agree with anyway.
I mean on TLPD Puzzle, who hasn't played for months, is still higher than Soulkey. And tear, who just made Code A, is higher than many Code S players.
How would such a list be possible anyway?
ELO
Completely remove any results by any player who hasn't played more than 3 sets in the last 2 months, and any games played prior to 4 months. Exclude any results from a non-major tournament. Only include players with at least 6 sets played. Pretty simple, really.
On February 02 2013 05:14 FuzzyJAM wrote: Without wishing to be a dick, this list is the best illustration that your stats are not close to being a good indicator of overall skill.
There isn't really a skill list achievable through statistics out there that people would agree with anyway.
I mean on TLPD Puzzle, who hasn't played for months, is still higher than Soulkey. And tear, who just made Code A, is higher than many Code S players.
How would such a list be possible anyway?
ELO
Completely remove any results by any player who hasn't played in the last 2 months. Exclude any results from a non-major tournament. Pretty simple, really.
What about someone who only plays a single game every two months?
LOL, I saw the discussion in the Code S LR thread - not too surprised there is now an official thread to discuss what Code S level means...
At least the OP is full with interesting numbers and analysis, always a good thing
In my personal opinion 'Code S Level' means 'someone who could do well in Code S', so someone who is close to being the best in the world. What this actually means in terms of skill, I don't know either. My analysis is definitely not 'Code S Level'
I personally think code s level is a expectation of skill level. This of course is a very broad term and i think not only dependent on wins or losses, but also on level of play displayed. In any game, anyone can lose against anyone. However, the way someone wins or loses says something about skill level (or relative skill level).
The definitions are highly subjective, as well as most of the elements in the definition (like in my definition), so an exact definition is impossible to give. At this moment, i think the best measurable approximation of a definition is 'those who are thought of as code s level the most, are code s level'. That is a very pragmatic approach, and i'd say it has some weaknesses, like fan-favorites being overestimated, or that it is susceptible to some biases.
So basically, 'what does code s level mean?' is not a meaningful question on it's own.
On February 02 2013 07:41 Mista_Masta wrote: My analysis is definitely not 'Code S Level'
haha, i liked that.
On February 02 2013 05:14 FuzzyJAM wrote: Without wishing to be a dick, this list is the best illustration that your stats are not close to being a good indicator of overall skill.
I don't think this can be said on it's own. Firstly, in my impression the list is not meant as an indicator of skill level, but it rather gives the chance someone will win a match against someone with another given position on this list, given historic win/loss data, disregarding any other possible influence of the game (format, location, physical/mental health, etc). Furthermore, i think that code s does not give a good impression of skill level either. The format of the gsl is significantly different from other tournaments, focussing on preparing for matches and/or opponents (as opposed to lots of matches in a small timespan, like mlg). I do think that the level in code s is higher overall, but the format definitely favours players who are good at mindgames, analysing the opponent and devising new strategies. If you define code s to be the standard of skill level, then i agree, this list is not a good indicator (although it was better than i'd expect). However, know that when doing so, you're assuming certain values as much as any other would be made list. i do agree that this list is not a good indicator for skill level, strictly speaking no list will ever..
On February 02 2013 05:14 FuzzyJAM wrote: Without wishing to be a dick, this list is the best illustration that your stats are not close to being a good indicator of overall skill.
There isn't really a skill list achievable through statistics out there that people would agree with anyway.
I mean on TLPD Puzzle, who hasn't played for months, is still higher than Soulkey. And tear, who just made Code A, is higher than many Code S players.
How would such a list be possible anyway?
ELO
Completely remove any results by any player who hasn't played in the last 2 months. Exclude any results from a non-major tournament. Pretty simple, really.
What about someone who only plays a single game every two months?
You do the same as in Chess ELO.
There is a certain time period before you become inactive, and you need a certain number of games before your rating is set in stone. Also, people whose ratings aren't set in stone count very little towards the ratings of others.
Therefor we don't get ridiculous results with people having no games having an absurd rating effecting the entire ladder, and all GSL players basically get placed appropriately.
You also don't have random foreigners in MLGs and other events completely destroying the ratings of top end koreans because they have a single bad set out of 10.
People who become inactive have their rating completely restarted from the beginning without effecting the games they've already played.
On February 02 2013 05:14 FuzzyJAM wrote: Without wishing to be a dick, this list is the best illustration that your stats are not close to being a good indicator of overall skill.
There isn't really a skill list achievable through statistics out there that people would agree with anyway.
I mean on TLPD Puzzle, who hasn't played for months, is still higher than Soulkey. And tear, who just made Code A, is higher than many Code S players.
How would such a list be possible anyway?
ELO
Completely remove any results by any player who hasn't played more than 3 sets in the last 2 months, and any games played prior to 4 months. Exclude any results from a non-major tournament. Only include players with at least 6 sets played. Pretty simple, really.
Wait wait wait..
You bash this system, and as a suggestion on what to use instead you say.. ELO? That's just.. wow. That is seriously amusing.
You do know that TLPD's rating is based on ELO, right?
On February 02 2013 00:18 Hiea wrote: The problem with this "ranking" is that some players play against really really high level players all the time, while some don't and others don't really ever play against the top, but are still higher.
Yes i agree. Taking the current 32 Code S Paricipants as a benchmark is pretty random imo and leads to some strange rankings like Scarlett far above Taeja for example. The benchmark should be something like what is a average win/loss rate in combination with the amount of appearences in Code S. Because being Code S-Level means to be able to repeat success in this league over a certain period of time.
Not what I meant, it's more that someone like Giant, has he ever competed against a korean? I think he took a game off Nerchio, but besides that...
Why does he rank higher than someone like Mvp? It shows that there is gigantic flaws in this list, and if I ever see it used as fact somewhere I will die a little inside.
It is what it is. It's a fact that this is what you get when you run the numbers the way I did. Anything outside of that is extrapolation, which may or may not be of much use. Anyone should be careful when they use statistics, and this is no exception.
But please, has Giantt ever competed against a Korean? The list is on the fucking website. It's not rocket science to actually go there and check. http://aligulac.com/players/294/results/
He has an even score against Hyun and 2-0 against TheStC for example.
And can I just say, why would that be the defining characteristic? A passport doesn't make you good at StarCraft. You should be asking about his results against highly rated players, not against Koreans.
He does NOT rank higher than Mvp. See here. They have almost the same rating (Mvp slightly higher). Like I have explanied, Giantt ranks higher than Mvp on the list given in the OP because it uses a different ranking system (performance against a pool of mostly Zergs and Terrans, matchups in which Giantt plays well).
But still, Mvp is almost always pointed out as a flaw in the system and I'm getting kinda tired of it. He hardly plays any more and his recent results are middling at best. http://aligulac.com/players/13/results/
He's just not very good and that's why he's not rated highly. Get over it.
Of course, as "code S level" goes, that can mean anything. The OP takes it to mean "would probably do ok in this season of code S", but that shows only one side of the story.
The real question people is asking, is why was Giantt adjusted to 2301 rating vs Terran dispite never facing someone above 2k?
He got a massive artificial boost without facing any Terrans even close to that rating (And only beating one, TheSTC 2-0 who wasn't even 2k rated)
Why was First artifically boosted FIVE HUNDRED rating in all match ups dispite facing no one even close to that skill?
http://aligulac.com/players/39/period/76/ How does this extrapolation make any sense in any world? You boosted him to the 2nd best player on the planet dispite him not even meeting anyone in the top 25.
In ELO you cannot beat on someone who is lower than 400 points below you for a REASON. Because utterly destroying a random master leaguer 100 times in a row does not make you the best in the world regardless of your expected results against them.
It's pretty simple and you know the answer yourself already. In this rating system you can get any rating no matter the ratings of your opponents, as long as you score well enough against them. In this respect it is different from Elo (not ELO). I can try to experiment with such a restriction, but I'm kinda cramped for time at the moment.
Again though, you can ease off on the attitude. I'm trying to be reasonable but you're a pain in the ass to talk to.
First was boosted five hundred points because it was 1000 points up to the maximal likelihood rating, and his past rating had similar variability to the ML rating, thus landing it somewhere in the middle. I didn't personally look at it and decide, y'know.
Fundamentally, I think one issue that most people neglect is what is the cut-off for "Code S level", if you are talking about the current code S, you are talking about 32 people. If you are taking about people who can do well in several code S's, that's probably about 16 people, or less. Without even going into deeper meaning of things, the number of people in the cut-off itself is very different. I mean, let's face it, if Stephano drew 3 players on Hack's level (given how he plays today), he will probably make it out on the top of the group.
On February 02 2013 08:07 achan1058 wrote: Fundamentally, I think one issue that most people neglect is what is the cut-off for "Code S level", if you are talking about the current code S, you are talking about 32 people. If you are taking about people who can do well in several code S's, that's probably about 16 people, or less. Without even going into deeper meaning of things, the number of people in the cut-off itself is very different. I mean, let's face it, if Stephano drew 3 players on Hack's level (given how he plays today), he will probably make it out on the top of the group.
It would be impossible for him to draw 3 players on Hack's level because of how the Code S group system works. You always have 1 person who got top 8 last season and then it is done by tiers with points. No offense to Hack he is very good but he is not ro8 level. Every player that got ro8 last season is better than him.
Stephano really didn't get a very hard group. Bogus looked great last GSL but has done very poorly in SPL. DRG is really good. Hack is pretty meh. At best it was a middle of the road group for this season. I'd say without a doubt groups A, B, C and E were harder.
On February 02 2013 01:17 Grumbels wrote: I think there are far too many 'Code S level' players out there. In Tennis, out of the last 40 premier tournaments, there are only five players that have won. The second place finishes are mostly these same players also. The number of players that are legitimately competitive in any given era is about a dozen. In Starcraft II it's rare to repeat a victory, it's even not unusual for the Ro8 to completely differ from tournament to tournament. I think there are about a hundred players that all can take Bo3's off each other. If you look at the full list in the OP, there are just so many good players out there, it's kind of insane.
I won't say it's always like this: there have been times where especially high level zergs were really stable and dominant, for instance DRG, Symbol, Leenock and Life have all had runs where they seemed unstoppable. MVP, MMA, Taeja have had similar runs. And that's a good thing to have for the game, but I don't think it happens often enough, and those players are still vulnerable. I don't think we'll ever see the same amount of consistency that Flash and Jaedong could achieve in Brood War. (unless Blizzard changes the game) I think in the future we will continue to see one-hit wonders.
You forgot wc3, ~80% of all premier tournaments were won by Grubby, Moon, ToD, Lyn or Sky since 2004 until sc2 release. I really miss this consistency in sc2. sc2:wol is really super random..
Agreed. The main problem with SC2:
1. There are more all ins and the risk:reward ratio is much higher than most games (more than BW).
2. A lot of matches mostly end in just one big engagement rather than through several smaller engagements like in BW or WC3 (this is a huge reason).
3. Due to the fog of war mechanic (which is present in most RTS), there is a luck factor involved. Due to the problem of #1 (all ins are more high risk and high reward) and #2 (most games involve just one big engagement, which means being caught out of position once can lose you the game), the luck and randomness factor increases more and thus SC2 becomes a more random game.
4. There are less different type of skills sets you can use in SC2 to win. Artosis pointed something interesting in SotG, talking about how there were some Zerg players who only went 2 hatch muta against all match ups, and won purely off of Mutalisk micro alone. Idra also said a while back (in 2011), that in BW, there were many different ways to outplay your opponent compared to SC2 (note - this isn't about Build Orders or anything, it's about certain mechanics or certain skills. The Mutalisk micro is a good example, it's not present in SC2, you can't win SC2 games in 'all match ups' off of going mutalisks alone; of course it might be a bit extreme example because the mutalisk is the best unit in BW and was viable in all 3 match ups but the point is that there are many different types of skills you can use in BW to win compared to SC2; a player who masters as many different types of skills as possible is likely to be a consistent champion; tl;dr version is the skill ceiling "isn't higher" in BW but the skill room in BW is "wider").
(Yes, skill room is a made up term but I think it's a good simple way to say that there are more types of skillsets you can use to win in BW compared to SC2. Still though, I definitely say #2, making the game more gradual instead of one battle into win/lose is most important to reduce random element in SC2.)
In BW, there were many ways and gameplay mechanics you could use to get wins. In SC2, there are also many ways to win but there are less. Generally, since everyone is doing more of the same as anyone else to win in SC2, this means that overlapping will happen more and there will be less ways to differentiate yourself from other types of players (again, not saying there isn't but it's much less compared to BW).
There's definitely way more competition in SC2 than a lot of other games though but the consistency of top level champions could be increased and randomness could be lowered if #2 and #4 were addressed.
On February 02 2013 08:07 achan1058 wrote: Fundamentally, I think one issue that most people neglect is what is the cut-off for "Code S level", if you are talking about the current code S, you are talking about 32 people. If you are taking about people who can do well in several code S's, that's probably about 16 people, or less. Without even going into deeper meaning of things, the number of people in the cut-off itself is very different. I mean, let's face it, if Stephano drew 3 players on Hack's level (given how he plays today), he will probably make it out on the top of the group.
It would be impossible for him to draw 3 players on Hack's level because of how the Code S group system works. You always have 1 person who got top 8 last season and then it is done by tiers with points. No offense to Hack he is very good but he is not ro8 level. Every player that got ro8 last season is better than him.
Stephano really didn't get a very hard group. Bogus looked great last GSL but has done very poorly in SPL. DRG is really good. Hack is pretty meh. At best it was a middle of the road group for this season. I'd say without a doubt groups A, B, C and E were harder.
My point still stands though. The difference in the number of players (hence the standard) between anyone who is currently in code S, compared to players consistently in code S, is huge, and not being careful with how you define it (not only in this thread, but in general discussions) will make a big difference.
Code S level mean exactly that, code S level. No need to analyze obvious things like what this community tends to do....great research but the energy and time could've been well spent elsewhere where it actually matters...
If you were to say a basketball team was a playoff caliber team, you would say they were good enough to make the playoffs, but either match-ups or consistency/streakiness would stand in their way. So doesn't this apply to Code S as well? Meaning that "Code S Level" is simply someone who is good enough to consistently compete with other Code S, or top level players, but may not currently be in Code S because of their Code A/B brackets giving match-up problems or a bad stretch of games causing them to drop out of Code S. To me there is no hard "S" line, but more of a moving target of skill levels/ranges
Seems pretty neat, but there are "code S" players who get knocked out of code S just because they had a bad day or unlucky group or something, so I don't think it's right for someone to take the highest ranked 32 players say they say the ones that aren't in code S belong there. It's certainly puts things in to perspective though.
In my mind, Code S players are very skilled at starcraft, but also are very skilled at preparing for matches. Maybe people underestimate the impact that preparation has in an RTS game. Many factors like team mates and practice partners play into this. So its not just the raw skill that makes a player a code S player
On February 02 2013 05:14 FuzzyJAM wrote: Without wishing to be a dick, this list is the best illustration that your stats are not close to being a good indicator of overall skill.
There isn't really a skill list achievable through statistics out there that people would agree with anyway.
I mean on TLPD Puzzle, who hasn't played for months, is still higher than Soulkey. And tear, who just made Code A, is higher than many Code S players.
How would such a list be possible anyway?
ELO
Completely remove any results by any player who hasn't played more than 3 sets in the last 2 months, and any games played prior to 4 months. Exclude any results from a non-major tournament. Only include players with at least 6 sets played. Pretty simple, really.
Wait wait wait..
You bash this system, and as a suggestion on what to use instead you say.. ELO? That's just.. wow. That is seriously amusing.
You do know that TLPD's rating is based on ELO, right?
You do know that TLPD counts every game and has no system for decay..... right? It's nothing like Chess ELO.
On February 02 2013 14:13 Burns wrote: In my mind, Code S players are very skilled at starcraft, but also are very skilled at preparing for matches. Maybe people underestimate the impact that preparation has in an RTS game. Many factors like team mates and practice partners play into this. So its not just the raw skill that makes a player a code S player
Yeah, if you put a lot of the current code S players in a bad environment for a long enough time and ask them to stay in code S, they'll have a lot of difficulties doing so. MC was doing relatively poorly before he started training in a team house again for example!
If you can win a game in Code S then you are Code S Level. That's it. Obviously people have ups and downs in form but being able to win in Code S is the strict definition. Some people get confused between current form and actual ability though.
I think what's cool and great about this work is that it does what statistics are good for. They give you the ability to create data and then to look at it critically. But maybe people in this thread confuse statistics with the Truth with a capital p. That's not the way you should read statistics, whether in the morning paper or on TL. Rather statistics can make us think about deeper connections that we haven't seen before, twist and turn around concepts and play with them through statistical models. They're not meant to say "Scarlett should be in Code S." Obviously there are flaws or issues with these stats, but it's common for statistics everywhere. What you can do with that is to add or change some modifier, let it meet other forms of reasoning or to extrapolate on what we take for granted.
So yeah, tl;dr, lies, damn lies and statistics are the case with all stats but it's not the point of stats.
On February 03 2013 02:50 Heartland wrote: I think what's cool and great about this work is that it does what statistics are good for. They give you the ability to create data and then to look at it critically. But maybe people in this thread confuse statistics with the Truth with a capital p. That's not the way you should read statistics, whether in the morning paper or on TL. Rather statistics can make us think about deeper connections that we haven't seen before, twist and turn around concepts and play with them through statistical models. They're not meant to say "Scarlett should be in Code S." Obviously there are flaws or issues with these stats, but it's common for statistics everywhere. What you can do with that is to add or change some modifier, let it meet other forms of reasoning or to extrapolate on what we take for granted.
So yeah, tl;dr, lies, damn lies and statistics are the case with all stats but it's not the point of stats.
TheBB, a great man who graced a small forum I frequent .
Good OP. Very creative mathematical model. It obviously relies a lot on past data rather than future expected outcomes, but it's really informative regardless. It's sad to see the BW greats ranked so low. They still are getting accustomed to the new game, but also SC2 is too different a game and depends more on chance than straight up mechanics and strategy, which is unfortunate.
On February 02 2013 22:28 revel8 wrote: If you can win a game in Code S then you are Code S Level. That's it. Obviously people have ups and downs in form but being able to win in Code S is the strict definition. Some people get confused between current form and actual ability though.
That means almost every known pro is Code S level... There would be like 400-500 players that can do that
On February 03 2013 04:53 JudicatorHammurabi wrote: TheBB, a great man who graced a small forum I frequent .
Huuummm... .org network?
Yes indeed. You know who sd bolts is. When I registered on TL, I got frustrated when the first 10 usernames I tried were already taken, and went with this one haha.
On February 02 2013 22:28 revel8 wrote: If you can win a game in Code S then you are Code S Level. That's it. Obviously people have ups and downs in form but being able to win in Code S is the strict definition. Some people get confused between current form and actual ability though.
That means almost every known pro is Code S level... There would be like 400-500 players that can do that
Well it is not about players who CAN win a game, but rather players who HAVE won a game in Code S. That is the strict definition of Code S Level. Of course there is a problem where people try and claim that Code S Level is a measure of a Player's SC2 ability. A player can lose a game in Code S and be eliminated. A player can win a game and be promoted into Code S. Although their status changes from being in Code S to not being in Code S or the opposite, their SC2 ability clearly does not change over the course of one series.
That is my opinion anyway, the term Code S Level does seem to be very subjective though.
EPT
Life 22.669 (code S)
First 21.351
Bomber 18.868 (code S)
you have parting as a zerg in your post, better change that
