EPTOn June 06 2026 02:28 RJBTVYOUTUBE wrote:
Should we hold Leta and YSC in higher regard than lets say Mini, who did not even qualify, even though Leta and YSC generally lose vs most pros and this ASL was more if a freak accident for them, albeit they did perform well, it was evidently way above their average if comparing it to their respective data sets including their usual opponents.
Both YSC and Leta for example went right back to being punching bags for most top and mid tier pros right after and even during ASL
Show nested quote +
On June 06 2026 02:22 Xenixx wrote:
That's fine, I'm not familiar with eloboard and never used it, but a lot of you are under the very mistaken idea(s) that you always get a representative sample size and that a statistic itself is inherently flawless. You very often don't, especially when you're not commissioning a study. You're quoting too much classroom statistics, not how it's often used in the real world. I understand the desire to want to use it like that but we do have limitations. And you have to work with what you have available.
One of the obvious variables here is including Proleague vs ASL. These are two very different competitions. Different format, often different players, different rules. Players often approach them differently than a prep-tournament, I know I did when I played this game back in the day.
I hold ASL players, especially past the Ro24, the top players, in far higher regard than anything else.
I don't see how any of that is unreasonable.
The question I'm sort of posing is what is more important for the topic of map balance, how the actual game(s) themselves were played or a compiled statistic of all games played without controlling for any variables. And if you're of the latter mindset then we will just have to agree to disagree but you are dead wrong if you think I'm the one misusing or misrepresenting the stats. I have a degree related to this and decades of experience in collection, analysis and compilation. I'm starting to resent the insinuation that you know more about this than I do when you're all making basic mistakes.
On June 05 2026 23:54 RJBTVYOUTUBE wrote:
You seem to misunderstand the data requirements/qualifications for Eloboard documentatoon/data inclusion.
1). can not be ladder
2). has to be livestreamed by both players if NOT an official tournament that broascasts/live streams it.
3). has to involve money, either from both players, from viewers, from tournament organizers, from sponsor organizations.
If a game does not meet the above requirements, the game will NOT be documented on Eloboard and NOT be included in the data set.
In addition, the skill difference between sometimes ASL qualified players, to ASL regulars, is much smaller than assumed.
About sample size: In small sample sizes variance in single data entries have a much greater effect on the total outcome of all entries due to the individual entry representing a larger percentage of the whole data set. 1/20 = 5% whereas 1/200 = 0.5%. Going by this fundamental aspect we get more reliable data intepretation the larger the number of entries in a data set is, because of replicability across similar data entries. This accounts for build order variance too.
If we have 10 games with 14cc vs zerg and the outcome is 60% win for zerg, we are forced to look at each individual game due variance in in-game actions and events having a much greater impact on the outcome. But if 14cc vs zerg has 56% over 100 games, we come much closer to the actual average outcome of the build order. And if we get to 55% at 1000 then that is the most likely most accurate representation.
On June 05 2026 22:48 Xenixx wrote:
And th For 862+601 games, who are you looking at? Ladder players with a bar code? Up and coming pro players? We all know there is such a massive gap between qualifier players for the ASL and ASL regulars.
If PoleStar is such a Terran favored map, one begs the question why Terrans couldn't win a game on it the previous two ASL seasons. If there's a clear bias or favoritism in the maps, you would expect it to manifest constantly or consistently. What we need, for a statistics alone approach, is tracked by build (roughly). Or broken down by meta (year?) and filtered through versions.
This is always the problem with statistics. Terrans would have a much lower winrate on average if they all suddenly decided to 14CC vs Z. But if you're suggesting that what happened in the individual games don't matter then there is a massive hole in your position here.
On June 05 2026 14:43 FlaShFTW wrote:
ELOBoard is the defacto map tracker as the people on there track all the sponmatches, ASL, proleague, etc. It's extremely well kept and keeps very good stats.
Also, because FlaSh doesn't even play that much online like sponmatches or proleague, these stats aren't going to be influenced heavily by removing him. He's not like an online monster that's crushing everyone, so if you wanted to remove FlaSh's stats, you'd have to think about removing Soulkey and Snow's stats for their respective races too, as Snow is known for being an online beast.
10% either way is a massive swing. 55-45 is a very strong map for the favored side. To keep things balanced, the most optimal maps try to keep it within 53-47 or 52-48. Maps like Radeon succeed here, and plenty of 2p and 3p maps which keep it in line for every single matchup.
You're looking at extremely small sample sizes based on individual tournaments, which we cannot derive any meaningful data from. Pole Star on ELOBoard has a TZ winrate of 862-601 or 58.9% Terran favored. Why would I use a few games from an isolated tournament compared to 1500 games played? This reasoning doesn't make any sense.
Look at the sample size. It's massive for the most played maps. Of course new maps like Octagon and Attitude will have lower sample sizes, but we've established a pattern as well of consistent trends with rotational symmetry 4p maps. It's all there mate. But there's no reason to be "controlling for the various variables" because all the variables work both ways. Better Terrans play weaker Zergs, and better Zergs play weaker Terrans. If you're going to try to argue against tens of thousands of games and a historical pattern since these stats have been getting tracked, you're going to need to put up a better argument than wasting people's time saying you need to look at every single individual game lol. But sure, be my guest, make it a little side project for you.
On June 05 2026 09:45 Xenixx wrote:
Are there better stats on maps available than what TL tracks? Someone may be able to narrow this down further with map versions in mind and I'm not going into tremendous depth here, I'll admit.
As always, when Flash plays, he inflates the win rates of maps. What if we cut his matches out of the stats. And beyond Snow, did any Protoss play well in the ASL? Bisu got the group of death, and Tulbo, well, yeah.
With what we got, if you're looking at these with statistics in mind <10% either way is well within the margin for error. Though I agree, anything close to 60% in any matchup indicates a possible favoritism. Are we considering 2-4% either way a bias towards a certain race? I don't. That seems rife with problems.
If you go back and look at the previous two seasons, 19 and 20, Pole Star, for example Terrans had a 0 and 16.7 percent win rate against Zergies. And if you go back to 19 and look at TvZ on a map like Death Valley (68% T) , Terrans didn't win a single game on it. Terrans pulled out an 83% win rate against Z on Octagon this season despite an overall (182 games tracked) 54% win rate. Roaring Currents is such a good example from last season about this too, did anyone who lost (Bisu, for example) play well on their ASL matches?
You get to a point, I think, that you have to look at the individual matches and when you do these statistics alone just don't hold up. It's exactly like the supposed gender pay gap in the US, when you start controlling for the various variables, it almost entirely disappears.
Are there better stats on maps available than what TL tracks? Someone may be able to narrow this down further with map versions in mind and I'm not going into tremendous depth here, I'll admit.
As always, when Flash plays, he inflates the win rates of maps. What if we cut his matches out of the stats. And beyond Snow, did any Protoss play well in the ASL? Bisu got the group of death, and Tulbo, well, yeah.
With what we got, if you're looking at these with statistics in mind <10% either way is well within the margin for error. Though I agree, anything close to 60% in any matchup indicates a possible favoritism. Are we considering 2-4% either way a bias towards a certain race? I don't. That seems rife with problems.
If you go back and look at the previous two seasons, 19 and 20, Pole Star, for example Terrans had a 0 and 16.7 percent win rate against Zergies. And if you go back to 19 and look at TvZ on a map like Death Valley (68% T) , Terrans didn't win a single game on it. Terrans pulled out an 83% win rate against Z on Octagon this season despite an overall (182 games tracked) 54% win rate. Roaring Currents is such a good example from last season about this too, did anyone who lost (Bisu, for example) play well on their ASL matches?
You get to a point, I think, that you have to look at the individual matches and when you do these statistics alone just don't hold up. It's exactly like the supposed gender pay gap in the US, when you start controlling for the various variables, it almost entirely disappears.
ELOBoard is the defacto map tracker as the people on there track all the sponmatches, ASL, proleague, etc. It's extremely well kept and keeps very good stats.
Also, because FlaSh doesn't even play that much online like sponmatches or proleague, these stats aren't going to be influenced heavily by removing him. He's not like an online monster that's crushing everyone, so if you wanted to remove FlaSh's stats, you'd have to think about removing Soulkey and Snow's stats for their respective races too, as Snow is known for being an online beast.
10% either way is a massive swing. 55-45 is a very strong map for the favored side. To keep things balanced, the most optimal maps try to keep it within 53-47 or 52-48. Maps like Radeon succeed here, and plenty of 2p and 3p maps which keep it in line for every single matchup.
You're looking at extremely small sample sizes based on individual tournaments, which we cannot derive any meaningful data from. Pole Star on ELOBoard has a TZ winrate of 862-601 or 58.9% Terran favored. Why would I use a few games from an isolated tournament compared to 1500 games played? This reasoning doesn't make any sense.
Look at the sample size. It's massive for the most played maps. Of course new maps like Octagon and Attitude will have lower sample sizes, but we've established a pattern as well of consistent trends with rotational symmetry 4p maps. It's all there mate. But there's no reason to be "controlling for the various variables" because all the variables work both ways. Better Terrans play weaker Zergs, and better Zergs play weaker Terrans. If you're going to try to argue against tens of thousands of games and a historical pattern since these stats have been getting tracked, you're going to need to put up a better argument than wasting people's time saying you need to look at every single individual game lol. But sure, be my guest, make it a little side project for you.
And th For 862+601 games, who are you looking at? Ladder players with a bar code? Up and coming pro players? We all know there is such a massive gap between qualifier players for the ASL and ASL regulars.
If PoleStar is such a Terran favored map, one begs the question why Terrans couldn't win a game on it the previous two ASL seasons. If there's a clear bias or favoritism in the maps, you would expect it to manifest constantly or consistently. What we need, for a statistics alone approach, is tracked by build (roughly). Or broken down by meta (year?) and filtered through versions.
This is always the problem with statistics. Terrans would have a much lower winrate on average if they all suddenly decided to 14CC vs Z. But if you're suggesting that what happened in the individual games don't matter then there is a massive hole in your position here.
You seem to misunderstand the data requirements/qualifications for Eloboard documentatoon/data inclusion.
1). can not be ladder
2). has to be livestreamed by both players if NOT an official tournament that broascasts/live streams it.
3). has to involve money, either from both players, from viewers, from tournament organizers, from sponsor organizations.
If a game does not meet the above requirements, the game will NOT be documented on Eloboard and NOT be included in the data set.
In addition, the skill difference between sometimes ASL qualified players, to ASL regulars, is much smaller than assumed.
About sample size: In small sample sizes variance in single data entries have a much greater effect on the total outcome of all entries due to the individual entry representing a larger percentage of the whole data set. 1/20 = 5% whereas 1/200 = 0.5%. Going by this fundamental aspect we get more reliable data intepretation the larger the number of entries in a data set is, because of replicability across similar data entries. This accounts for build order variance too.
If we have 10 games with 14cc vs zerg and the outcome is 60% win for zerg, we are forced to look at each individual game due variance in in-game actions and events having a much greater impact on the outcome. But if 14cc vs zerg has 56% over 100 games, we come much closer to the actual average outcome of the build order. And if we get to 55% at 1000 then that is the most likely most accurate representation.
That's fine, I'm not familiar with eloboard and never used it, but a lot of you are under the very mistaken idea(s) that you always get a representative sample size and that a statistic itself is inherently flawless. You very often don't, especially when you're not commissioning a study. You're quoting too much classroom statistics, not how it's often used in the real world. I understand the desire to want to use it like that but we do have limitations. And you have to work with what you have available.
One of the obvious variables here is including Proleague vs ASL. These are two very different competitions. Different format, often different players, different rules. Players often approach them differently than a prep-tournament, I know I did when I played this game back in the day.
I hold ASL players, especially past the Ro24, the top players, in far higher regard than anything else.
I don't see how any of that is unreasonable.
The question I'm sort of posing is what is more important for the topic of map balance, how the actual game(s) themselves were played or a compiled statistic of all games played without controlling for any variables. And if you're of the latter mindset then we will just have to agree to disagree but you are dead wrong if you think I'm the one misusing or misrepresenting the stats. I have a degree related to this and decades of experience in collection, analysis and compilation. I'm starting to resent the insinuation that you know more about this than I do when you're all making basic mistakes.
Should we hold Leta and YSC in higher regard than lets say Mini, who did not even qualify, even though Leta and YSC generally lose vs most pros and this ASL was more if a freak accident for them, albeit they did perform well, it was evidently way above their average if comparing it to their respective data sets including their usual opponents.
Both YSC and Leta for example went right back to being punching bags for most top and mid tier pros right after and even during ASL
Wrong question, but at least you're heading in the right direction. The question is, are the maps they played on biased or was it due to other variables like, how they played?
![[image loading]](https://i.imgur.com/y53aezc.png)
![[image loading]](https://i.imgur.com/2hJYrmm.png)
![[image loading]](https://i.imgur.com/HvqpIOy.png)
![[image loading]](https://i.imgur.com/EjTIx6h.png)
![[image loading]](https://i.imgur.com/tiCl7sP.png)
![[image loading]](https://i.imgur.com/ef903Xq.png)
![[image loading]](https://i.imgur.com/4tbdNqW.png)
![[image loading]](https://i.imgur.com/5mjmU5w.png)
![[image loading]](https://i.imgur.com/r9BmoJX.png)
