|
On February 13 2014 05:13 Destructicon wrote:Show nested quote +On February 13 2014 04:56 DinoMight wrote:On February 13 2014 04:43 Destructicon wrote:
Well, due to imbalance what turns out happening in tournaments is this.
A lot of weaker representatives of their race qualify for a tournament due to their race being stronger at a time.
That's sign number one of imbalance, over representation in qualifiers if all else is equal.
Second thing that happens is that, in the lower parts of the main tournaments the masses of bad players that where carried by imbalance get slowly culled by the way better players of the other race, giving the false perception of balance by skewing win rates closer to 50%. This is another sign of imbalance but a much more subtle one.
Lastly when players of equal skill meet up in the higher up parts of the tournament then the imbalanced race starts dominating again and they occupy most of the podium spots. This is the last sign of imbalance.
You can observe all these signs of imbalance in all the previous periods of massive domination from one race or another. So i guess you could look at like... race ratio in first round / qualifier vs. race ratio in Ro16/8 vs. podium finishes. I don't know if there is enough data for something like this to be statistically significant though (in the academic definition of significant). You'd need a ton of tournaments and players. You're right though, tournament win rates are kind of silly to look at for balance. Because Tajea is going to beat a mediocre Protoss regardless of whether or not the game is balanced. Taeja can even beat some top tier Protosses despite any imbalance that may exist, because he's really really good. I think they should just take (non mirror wins for race x) / (non mirror games played by race x) and literally just do this for every single ladder game played total. You can't do it by league because the leagues are skewing the data to 50% intentionally. Wouldn't that technically show balance or imbalance>? Actually no, about 4-5 tournaments over a period of 3 months is enough data usually. I'd say IEM Sao Paulo, ASUS RoG, and the Code B and Code A pooled data is nearly enough, Code S finishing would be even more conclusive evidence. WCS EU is kind of poor because we have a huge mix of different skill levels because Koreans out match most of the EU players. And pooling all ladder data is useless since you'll see tons of low level plays. Data should only be collected from Masters and GM, Korean qualifiers or any qualifiers with lots of Koreans involved and premier tournaments with a lot of Koreans. Edit: And again, by data I mean a combination of win rates and race distribution across ladder and tournaments.
But if you're only selecting Masters and GM then you run into the same problem you were just discussing where the ladder forces a 50% winrate regardless of how good you actually are. How many people in GM actually have a win rate statistically > 50%?? I also don't think you should be looking at only Koreans.
So I think maybe if you looked at winrate averages for those players who are good enough to be statistically above the 50% that Blizzard forces you to you could get an idea of racial balance. If all of them are Protoss or Terran or Zerg then you'd know that the only way to be better than everyone else is to play this race. Well, you wouldn't know for sure but it would be likely.
|
So something like 8% of Protoss in GM have >=55% winrate. 6% of Terrans, and 4% of Zerg.
But that's really pushing it. Not enough data to do this IMO.
|
On February 13 2014 05:29 DinoMight wrote:Show nested quote +On February 13 2014 05:13 Destructicon wrote:On February 13 2014 04:56 DinoMight wrote:On February 13 2014 04:43 Destructicon wrote:
Well, due to imbalance what turns out happening in tournaments is this.
A lot of weaker representatives of their race qualify for a tournament due to their race being stronger at a time.
That's sign number one of imbalance, over representation in qualifiers if all else is equal.
Second thing that happens is that, in the lower parts of the main tournaments the masses of bad players that where carried by imbalance get slowly culled by the way better players of the other race, giving the false perception of balance by skewing win rates closer to 50%. This is another sign of imbalance but a much more subtle one.
Lastly when players of equal skill meet up in the higher up parts of the tournament then the imbalanced race starts dominating again and they occupy most of the podium spots. This is the last sign of imbalance.
You can observe all these signs of imbalance in all the previous periods of massive domination from one race or another. So i guess you could look at like... race ratio in first round / qualifier vs. race ratio in Ro16/8 vs. podium finishes. I don't know if there is enough data for something like this to be statistically significant though (in the academic definition of significant). You'd need a ton of tournaments and players. You're right though, tournament win rates are kind of silly to look at for balance. Because Tajea is going to beat a mediocre Protoss regardless of whether or not the game is balanced. Taeja can even beat some top tier Protosses despite any imbalance that may exist, because he's really really good. I think they should just take (non mirror wins for race x) / (non mirror games played by race x) and literally just do this for every single ladder game played total. You can't do it by league because the leagues are skewing the data to 50% intentionally. Wouldn't that technically show balance or imbalance>? Actually no, about 4-5 tournaments over a period of 3 months is enough data usually. I'd say IEM Sao Paulo, ASUS RoG, and the Code B and Code A pooled data is nearly enough, Code S finishing would be even more conclusive evidence. WCS EU is kind of poor because we have a huge mix of different skill levels because Koreans out match most of the EU players. And pooling all ladder data is useless since you'll see tons of low level plays. Data should only be collected from Masters and GM, Korean qualifiers or any qualifiers with lots of Koreans involved and premier tournaments with a lot of Koreans. Edit: And again, by data I mean a combination of win rates and race distribution across ladder and tournaments. But if you're only selecting Masters and GM then you run into the same problem you were just discussing where the ladder forces a 50% winrate regardless of how good you actually are. How many people in GM actually have a win rate statistically > 50%?? I also don't think you should be looking at only Koreans. So I think maybe if you looked at winrate averages for those players who are good enough to be statistically above the 50% that Blizzard forces you to you could get an idea of racial balance. If all of them are Protoss or Terran or Zerg then you'd know that the only way to be better than everyone else is to play this race. Well, you wouldn't know for sure but it would be likely.
Ladder doesn't force a 50% winrate. If P >> T, but all the other match-ups are balanced, this will show up in the ladder data that Blizzard gets. In fact, it will show that Z > T, because the zergs tend to be stronger as their MMR will be lower due to playing protoss. The only time it won't show up is when, say, terran is equally bad against both zerg and protoss.
Also, Blizzard's adjusted win percentages are a bit more accurate than pure ladder data, although I think the same rules still apply.
GM+Masters race representation is probably more meaningful, but you'd have to take into account the number of players of each race and various intangible factors such as that players of some races are more likely to go pro.
And ladder data can't take into account that protoss might be better when you are playing vs barcodes but worse when playing vs known opponents.
|
On February 13 2014 05:29 DinoMight wrote:Show nested quote +On February 13 2014 05:13 Destructicon wrote:On February 13 2014 04:56 DinoMight wrote:On February 13 2014 04:43 Destructicon wrote:
Well, due to imbalance what turns out happening in tournaments is this.
A lot of weaker representatives of their race qualify for a tournament due to their race being stronger at a time.
That's sign number one of imbalance, over representation in qualifiers if all else is equal.
Second thing that happens is that, in the lower parts of the main tournaments the masses of bad players that where carried by imbalance get slowly culled by the way better players of the other race, giving the false perception of balance by skewing win rates closer to 50%. This is another sign of imbalance but a much more subtle one.
Lastly when players of equal skill meet up in the higher up parts of the tournament then the imbalanced race starts dominating again and they occupy most of the podium spots. This is the last sign of imbalance.
You can observe all these signs of imbalance in all the previous periods of massive domination from one race or another. So i guess you could look at like... race ratio in first round / qualifier vs. race ratio in Ro16/8 vs. podium finishes. I don't know if there is enough data for something like this to be statistically significant though (in the academic definition of significant). You'd need a ton of tournaments and players. You're right though, tournament win rates are kind of silly to look at for balance. Because Tajea is going to beat a mediocre Protoss regardless of whether or not the game is balanced. Taeja can even beat some top tier Protosses despite any imbalance that may exist, because he's really really good. I think they should just take (non mirror wins for race x) / (non mirror games played by race x) and literally just do this for every single ladder game played total. You can't do it by league because the leagues are skewing the data to 50% intentionally. Wouldn't that technically show balance or imbalance>? Actually no, about 4-5 tournaments over a period of 3 months is enough data usually. I'd say IEM Sao Paulo, ASUS RoG, and the Code B and Code A pooled data is nearly enough, Code S finishing would be even more conclusive evidence. WCS EU is kind of poor because we have a huge mix of different skill levels because Koreans out match most of the EU players. And pooling all ladder data is useless since you'll see tons of low level plays. Data should only be collected from Masters and GM, Korean qualifiers or any qualifiers with lots of Koreans involved and premier tournaments with a lot of Koreans. Edit: And again, by data I mean a combination of win rates and race distribution across ladder and tournaments. But if you're only selecting Masters and GM then you run into the same problem you were just discussing where the ladder forces a 50% winrate regardless of how good you actually are. How many people in GM actually have a win rate statistically > 50%?? I also don't think you should be looking at only Koreans. So I think maybe if you looked at winrate averages for those players who are good enough to be statistically above the 50% that Blizzard forces you to you could get an idea of racial balance. If all of them are Protoss or Terran or Zerg then you'd know that the only way to be better than everyone else is to play this race. Well, you wouldn't know for sure but it would be likely.
Racial distribution in masters/GM says something.
Also, you can look at foreign pros, but they are all relatively 'bad' and will all lose to korean pros (except Scarlett, Naniwa and a select few), and thus scew numbers.
|
On February 13 2014 05:40 SC2Toastie wrote:Show nested quote +On February 13 2014 05:29 DinoMight wrote:On February 13 2014 05:13 Destructicon wrote:On February 13 2014 04:56 DinoMight wrote:On February 13 2014 04:43 Destructicon wrote:
Well, due to imbalance what turns out happening in tournaments is this.
A lot of weaker representatives of their race qualify for a tournament due to their race being stronger at a time.
That's sign number one of imbalance, over representation in qualifiers if all else is equal.
Second thing that happens is that, in the lower parts of the main tournaments the masses of bad players that where carried by imbalance get slowly culled by the way better players of the other race, giving the false perception of balance by skewing win rates closer to 50%. This is another sign of imbalance but a much more subtle one.
Lastly when players of equal skill meet up in the higher up parts of the tournament then the imbalanced race starts dominating again and they occupy most of the podium spots. This is the last sign of imbalance.
You can observe all these signs of imbalance in all the previous periods of massive domination from one race or another. So i guess you could look at like... race ratio in first round / qualifier vs. race ratio in Ro16/8 vs. podium finishes. I don't know if there is enough data for something like this to be statistically significant though (in the academic definition of significant). You'd need a ton of tournaments and players. You're right though, tournament win rates are kind of silly to look at for balance. Because Tajea is going to beat a mediocre Protoss regardless of whether or not the game is balanced. Taeja can even beat some top tier Protosses despite any imbalance that may exist, because he's really really good. I think they should just take (non mirror wins for race x) / (non mirror games played by race x) and literally just do this for every single ladder game played total. You can't do it by league because the leagues are skewing the data to 50% intentionally. Wouldn't that technically show balance or imbalance>? Actually no, about 4-5 tournaments over a period of 3 months is enough data usually. I'd say IEM Sao Paulo, ASUS RoG, and the Code B and Code A pooled data is nearly enough, Code S finishing would be even more conclusive evidence. WCS EU is kind of poor because we have a huge mix of different skill levels because Koreans out match most of the EU players. And pooling all ladder data is useless since you'll see tons of low level plays. Data should only be collected from Masters and GM, Korean qualifiers or any qualifiers with lots of Koreans involved and premier tournaments with a lot of Koreans. Edit: And again, by data I mean a combination of win rates and race distribution across ladder and tournaments. But if you're only selecting Masters and GM then you run into the same problem you were just discussing where the ladder forces a 50% winrate regardless of how good you actually are. How many people in GM actually have a win rate statistically > 50%?? I also don't think you should be looking at only Koreans. So I think maybe if you looked at winrate averages for those players who are good enough to be statistically above the 50% that Blizzard forces you to you could get an idea of racial balance. If all of them are Protoss or Terran or Zerg then you'd know that the only way to be better than everyone else is to play this race. Well, you wouldn't know for sure but it would be likely. Racial distribution in masters/GM says something. Also, you can look at foreign pros, but they are all relatively 'bad' and will all lose to korean pros (except Scarlett, Naniwa and a select few), and thus scew numbers.
Well, I think Protoss has the most to gain from practicing on ladder as they're the agressors early in the two non mirror matchups. So it makes sense that they would play a lot more ladder and therefore rank higher / in more numbers than Terran/Zerg. A lot of really really good Korean Terrans are nowhere to be found on the ladder.. From what I've heard a lot of the Terran/Zerg players practice in customs and don't play so much ladder.
It's easy to say Protoss imba because obviously they have the most players in Masters/GM but I think there are many other factors that play into this.. honestly I think global win rates would be, statistically, the best test. Sure it would be messy as hell, but all things equal, if you took every game of starcraft played (ladder AND custom) and did vanilla win rates, I see no reason for one race to have more win % than another unless there is imbalance.
|
On February 13 2014 05:59 DinoMight wrote:Show nested quote +On February 13 2014 05:40 SC2Toastie wrote:On February 13 2014 05:29 DinoMight wrote:On February 13 2014 05:13 Destructicon wrote:On February 13 2014 04:56 DinoMight wrote:On February 13 2014 04:43 Destructicon wrote:
Well, due to imbalance what turns out happening in tournaments is this.
A lot of weaker representatives of their race qualify for a tournament due to their race being stronger at a time.
That's sign number one of imbalance, over representation in qualifiers if all else is equal.
Second thing that happens is that, in the lower parts of the main tournaments the masses of bad players that where carried by imbalance get slowly culled by the way better players of the other race, giving the false perception of balance by skewing win rates closer to 50%. This is another sign of imbalance but a much more subtle one.
Lastly when players of equal skill meet up in the higher up parts of the tournament then the imbalanced race starts dominating again and they occupy most of the podium spots. This is the last sign of imbalance.
You can observe all these signs of imbalance in all the previous periods of massive domination from one race or another. So i guess you could look at like... race ratio in first round / qualifier vs. race ratio in Ro16/8 vs. podium finishes. I don't know if there is enough data for something like this to be statistically significant though (in the academic definition of significant). You'd need a ton of tournaments and players. You're right though, tournament win rates are kind of silly to look at for balance. Because Tajea is going to beat a mediocre Protoss regardless of whether or not the game is balanced. Taeja can even beat some top tier Protosses despite any imbalance that may exist, because he's really really good. I think they should just take (non mirror wins for race x) / (non mirror games played by race x) and literally just do this for every single ladder game played total. You can't do it by league because the leagues are skewing the data to 50% intentionally. Wouldn't that technically show balance or imbalance>? Actually no, about 4-5 tournaments over a period of 3 months is enough data usually. I'd say IEM Sao Paulo, ASUS RoG, and the Code B and Code A pooled data is nearly enough, Code S finishing would be even more conclusive evidence. WCS EU is kind of poor because we have a huge mix of different skill levels because Koreans out match most of the EU players. And pooling all ladder data is useless since you'll see tons of low level plays. Data should only be collected from Masters and GM, Korean qualifiers or any qualifiers with lots of Koreans involved and premier tournaments with a lot of Koreans. Edit: And again, by data I mean a combination of win rates and race distribution across ladder and tournaments. But if you're only selecting Masters and GM then you run into the same problem you were just discussing where the ladder forces a 50% winrate regardless of how good you actually are. How many people in GM actually have a win rate statistically > 50%?? I also don't think you should be looking at only Koreans. So I think maybe if you looked at winrate averages for those players who are good enough to be statistically above the 50% that Blizzard forces you to you could get an idea of racial balance. If all of them are Protoss or Terran or Zerg then you'd know that the only way to be better than everyone else is to play this race. Well, you wouldn't know for sure but it would be likely. Racial distribution in masters/GM says something. Also, you can look at foreign pros, but they are all relatively 'bad' and will all lose to korean pros (except Scarlett, Naniwa and a select few), and thus scew numbers. Well, I think Protoss has the most to gain from practicing on ladder as they're the agressors early in the two non mirror matchups. So it makes sense that they would play a lot more ladder and therefore rank higher / in more numbers than Terran/Zerg. A lot of really really good Korean Terrans are nowhere to be found on the ladder.. From what I've heard a lot of the Terran/Zerg players practice in customs and don't play so much ladder. It's easy to say Protoss imba because obviously they have the most players in Masters/GM but I think there are many other factors that play into this.. honestly I think global win rates would be, statistically, the best test. Sure it would be messy as hell, but all things equal, if you took every game of starcraft played (ladder AND custom) and did vanilla win rates, I see no reason for one race to have more win % than another unless there is imbalance.
?? Sorry, I don't get the sense of the logical connectives here. How exactly does "being the agressor early game" translates to "need more practice" ?...
Also how on earth would you know that "a lot of really really good Korean Terrans are nowhere to be found on the ladder" since 95% of the players are barcodes there?
|
On February 13 2014 05:38 DinoMight wrote:
So something like 8% of Protoss in GM have >=55% winrate. 6% of Terrans, and 4% of Zerg.
But that's really pushing it. Not enough data to do this IMO.
8% out of a much bigger pool, lets not forget...
|
4713 Posts
On February 13 2014 05:59 DinoMight wrote:Show nested quote +On February 13 2014 05:40 SC2Toastie wrote:On February 13 2014 05:29 DinoMight wrote:On February 13 2014 05:13 Destructicon wrote:On February 13 2014 04:56 DinoMight wrote:On February 13 2014 04:43 Destructicon wrote:
Well, due to imbalance what turns out happening in tournaments is this.
A lot of weaker representatives of their race qualify for a tournament due to their race being stronger at a time.
That's sign number one of imbalance, over representation in qualifiers if all else is equal.
Second thing that happens is that, in the lower parts of the main tournaments the masses of bad players that where carried by imbalance get slowly culled by the way better players of the other race, giving the false perception of balance by skewing win rates closer to 50%. This is another sign of imbalance but a much more subtle one.
Lastly when players of equal skill meet up in the higher up parts of the tournament then the imbalanced race starts dominating again and they occupy most of the podium spots. This is the last sign of imbalance.
You can observe all these signs of imbalance in all the previous periods of massive domination from one race or another. So i guess you could look at like... race ratio in first round / qualifier vs. race ratio in Ro16/8 vs. podium finishes. I don't know if there is enough data for something like this to be statistically significant though (in the academic definition of significant). You'd need a ton of tournaments and players. You're right though, tournament win rates are kind of silly to look at for balance. Because Tajea is going to beat a mediocre Protoss regardless of whether or not the game is balanced. Taeja can even beat some top tier Protosses despite any imbalance that may exist, because he's really really good. I think they should just take (non mirror wins for race x) / (non mirror games played by race x) and literally just do this for every single ladder game played total. You can't do it by league because the leagues are skewing the data to 50% intentionally. Wouldn't that technically show balance or imbalance>? Actually no, about 4-5 tournaments over a period of 3 months is enough data usually. I'd say IEM Sao Paulo, ASUS RoG, and the Code B and Code A pooled data is nearly enough, Code S finishing would be even more conclusive evidence. WCS EU is kind of poor because we have a huge mix of different skill levels because Koreans out match most of the EU players. And pooling all ladder data is useless since you'll see tons of low level plays. Data should only be collected from Masters and GM, Korean qualifiers or any qualifiers with lots of Koreans involved and premier tournaments with a lot of Koreans. Edit: And again, by data I mean a combination of win rates and race distribution across ladder and tournaments. But if you're only selecting Masters and GM then you run into the same problem you were just discussing where the ladder forces a 50% winrate regardless of how good you actually are. How many people in GM actually have a win rate statistically > 50%?? I also don't think you should be looking at only Koreans. So I think maybe if you looked at winrate averages for those players who are good enough to be statistically above the 50% that Blizzard forces you to you could get an idea of racial balance. If all of them are Protoss or Terran or Zerg then you'd know that the only way to be better than everyone else is to play this race. Well, you wouldn't know for sure but it would be likely. Racial distribution in masters/GM says something. Also, you can look at foreign pros, but they are all relatively 'bad' and will all lose to korean pros (except Scarlett, Naniwa and a select few), and thus scew numbers. Well, I think Protoss has the most to gain from practicing on ladder as they're the agressors early in the two non mirror matchups. So it makes sense that they would play a lot more ladder and therefore rank higher / in more numbers than Terran/Zerg. A lot of really really good Korean Terrans are nowhere to be found on the ladder.. From what I've heard a lot of the Terran/Zerg players practice in customs and don't play so much ladder. It's easy to say Protoss imba because obviously they have the most players in Masters/GM but I think there are many other factors that play into this.. honestly I think global win rates would be, statistically, the best test. Sure it would be messy as hell, but all things equal, if you took every game of starcraft played (ladder AND custom) and did vanilla win rates, I see no reason for one race to have more win % than another unless there is imbalance.
I'm quite sure that during the BL Infestor era the ladder in masters and GM was still dominated mostly by zergs, just like its now dominated by protoss in terms of racial distribution.
|
On February 13 2014 06:06 TheDwf wrote:Show nested quote +On February 13 2014 05:59 DinoMight wrote:On February 13 2014 05:40 SC2Toastie wrote:On February 13 2014 05:29 DinoMight wrote:On February 13 2014 05:13 Destructicon wrote:On February 13 2014 04:56 DinoMight wrote:On February 13 2014 04:43 Destructicon wrote:
Well, due to imbalance what turns out happening in tournaments is this.
A lot of weaker representatives of their race qualify for a tournament due to their race being stronger at a time.
That's sign number one of imbalance, over representation in qualifiers if all else is equal.
Second thing that happens is that, in the lower parts of the main tournaments the masses of bad players that where carried by imbalance get slowly culled by the way better players of the other race, giving the false perception of balance by skewing win rates closer to 50%. This is another sign of imbalance but a much more subtle one.
Lastly when players of equal skill meet up in the higher up parts of the tournament then the imbalanced race starts dominating again and they occupy most of the podium spots. This is the last sign of imbalance.
You can observe all these signs of imbalance in all the previous periods of massive domination from one race or another. So i guess you could look at like... race ratio in first round / qualifier vs. race ratio in Ro16/8 vs. podium finishes. I don't know if there is enough data for something like this to be statistically significant though (in the academic definition of significant). You'd need a ton of tournaments and players. You're right though, tournament win rates are kind of silly to look at for balance. Because Tajea is going to beat a mediocre Protoss regardless of whether or not the game is balanced. Taeja can even beat some top tier Protosses despite any imbalance that may exist, because he's really really good. I think they should just take (non mirror wins for race x) / (non mirror games played by race x) and literally just do this for every single ladder game played total. You can't do it by league because the leagues are skewing the data to 50% intentionally. Wouldn't that technically show balance or imbalance>? Actually no, about 4-5 tournaments over a period of 3 months is enough data usually. I'd say IEM Sao Paulo, ASUS RoG, and the Code B and Code A pooled data is nearly enough, Code S finishing would be even more conclusive evidence. WCS EU is kind of poor because we have a huge mix of different skill levels because Koreans out match most of the EU players. And pooling all ladder data is useless since you'll see tons of low level plays. Data should only be collected from Masters and GM, Korean qualifiers or any qualifiers with lots of Koreans involved and premier tournaments with a lot of Koreans. Edit: And again, by data I mean a combination of win rates and race distribution across ladder and tournaments. But if you're only selecting Masters and GM then you run into the same problem you were just discussing where the ladder forces a 50% winrate regardless of how good you actually are. How many people in GM actually have a win rate statistically > 50%?? I also don't think you should be looking at only Koreans. So I think maybe if you looked at winrate averages for those players who are good enough to be statistically above the 50% that Blizzard forces you to you could get an idea of racial balance. If all of them are Protoss or Terran or Zerg then you'd know that the only way to be better than everyone else is to play this race. Well, you wouldn't know for sure but it would be likely. Racial distribution in masters/GM says something. Also, you can look at foreign pros, but they are all relatively 'bad' and will all lose to korean pros (except Scarlett, Naniwa and a select few), and thus scew numbers. Well, I think Protoss has the most to gain from practicing on ladder as they're the agressors early in the two non mirror matchups. So it makes sense that they would play a lot more ladder and therefore rank higher / in more numbers than Terran/Zerg. A lot of really really good Korean Terrans are nowhere to be found on the ladder.. From what I've heard a lot of the Terran/Zerg players practice in customs and don't play so much ladder. It's easy to say Protoss imba because obviously they have the most players in Masters/GM but I think there are many other factors that play into this.. honestly I think global win rates would be, statistically, the best test. Sure it would be messy as hell, but all things equal, if you took every game of starcraft played (ladder AND custom) and did vanilla win rates, I see no reason for one race to have more win % than another unless there is imbalance. ?? Sorry, I don't get the sense of the logical connectives here. How exactly does "being the agressor early game" translates to "need more practice" ?... Also how on earth would you know that "a lot of really really good Korean Terrans are nowhere to be found on the ladder" since 95% of the players are barcodes there?
I'm precisely trying to avoid to jump to connections of causality, but one theory could be that as Protoss if you want to practice something you can just go on ladder and practice it. But as Terran and Zerg you could be practicing something and then get Oracle'd. Or Blink allined, or 4 gated.. etc. Protoss dictates the game in those early stages a lot. So ladder is less effective to practice something and therefore you wouldn't play as much on it.
The statements I made are not fact, just an opinion. I saw in interviews innovation and MMA say they play a lot of customs w/ Scarlett, but Dear said he mostly just plays ladder. Some other pros made similar comments in interviews. That's what I was going off of.
|
On February 13 2014 05:59 DinoMight wrote:Show nested quote +On February 13 2014 05:40 SC2Toastie wrote:On February 13 2014 05:29 DinoMight wrote:On February 13 2014 05:13 Destructicon wrote:On February 13 2014 04:56 DinoMight wrote:On February 13 2014 04:43 Destructicon wrote:
Well, due to imbalance what turns out happening in tournaments is this.
A lot of weaker representatives of their race qualify for a tournament due to their race being stronger at a time.
That's sign number one of imbalance, over representation in qualifiers if all else is equal.
Second thing that happens is that, in the lower parts of the main tournaments the masses of bad players that where carried by imbalance get slowly culled by the way better players of the other race, giving the false perception of balance by skewing win rates closer to 50%. This is another sign of imbalance but a much more subtle one.
Lastly when players of equal skill meet up in the higher up parts of the tournament then the imbalanced race starts dominating again and they occupy most of the podium spots. This is the last sign of imbalance.
You can observe all these signs of imbalance in all the previous periods of massive domination from one race or another. So i guess you could look at like... race ratio in first round / qualifier vs. race ratio in Ro16/8 vs. podium finishes. I don't know if there is enough data for something like this to be statistically significant though (in the academic definition of significant). You'd need a ton of tournaments and players. You're right though, tournament win rates are kind of silly to look at for balance. Because Tajea is going to beat a mediocre Protoss regardless of whether or not the game is balanced. Taeja can even beat some top tier Protosses despite any imbalance that may exist, because he's really really good. I think they should just take (non mirror wins for race x) / (non mirror games played by race x) and literally just do this for every single ladder game played total. You can't do it by league because the leagues are skewing the data to 50% intentionally. Wouldn't that technically show balance or imbalance>? Actually no, about 4-5 tournaments over a period of 3 months is enough data usually. I'd say IEM Sao Paulo, ASUS RoG, and the Code B and Code A pooled data is nearly enough, Code S finishing would be even more conclusive evidence. WCS EU is kind of poor because we have a huge mix of different skill levels because Koreans out match most of the EU players. And pooling all ladder data is useless since you'll see tons of low level plays. Data should only be collected from Masters and GM, Korean qualifiers or any qualifiers with lots of Koreans involved and premier tournaments with a lot of Koreans. Edit: And again, by data I mean a combination of win rates and race distribution across ladder and tournaments. But if you're only selecting Masters and GM then you run into the same problem you were just discussing where the ladder forces a 50% winrate regardless of how good you actually are. How many people in GM actually have a win rate statistically > 50%?? I also don't think you should be looking at only Koreans. So I think maybe if you looked at winrate averages for those players who are good enough to be statistically above the 50% that Blizzard forces you to you could get an idea of racial balance. If all of them are Protoss or Terran or Zerg then you'd know that the only way to be better than everyone else is to play this race. Well, you wouldn't know for sure but it would be likely. Racial distribution in masters/GM says something. Also, you can look at foreign pros, but they are all relatively 'bad' and will all lose to korean pros (except Scarlett, Naniwa and a select few), and thus scew numbers. Well, I think Protoss has the most to gain from practicing on ladder as they're the agressors early in the two non mirror matchups. So it makes sense that they would play a lot more ladder and therefore rank higher / in more numbers than Terran/Zerg. A lot of really really good Korean Terrans are nowhere to be found on the ladder.. From what I've heard a lot of the Terran/Zerg players practice in customs and don't play so much ladder. It's easy to say Protoss imba because obviously they have the most players in Masters/GM but I think there are many other factors that play into this.. honestly I think global win rates would be, statistically, the best test. Sure it would be messy as hell, but all things equal, if you took every game of starcraft played (ladder AND custom) and did vanilla win rates, I see no reason for one race to have more win % than another unless there is imbalance.
Does vanilla wn rate include bronze level match ? How is that useful. If you only include from certain levels up then run into same proplem.
First paragraph such trolling lol. You try so hard to defend protoss it makes you look silly.
|
On February 13 2014 06:17 Snusmumriken wrote:Show nested quote +On February 13 2014 05:38 DinoMight wrote:
So something like 8% of Protoss in GM have >=55% winrate. 6% of Terrans, and 4% of Zerg.
But that's really pushing it. Not enough data to do this IMO. 8% out of a much bigger pool, lets not forget...
Well, i'm trying to isolate imbalance issues and I dont really think the total pool really impacts things. For example if there was real imbalance in TvZ, you could have 100 Zergs but the one Terran would always win. You couldn't say there are more Zergs because Zerg is imbalanced for sure. Maybe other factors are playing into it.
So yeah jsut trying to be really scientific about finding a way to measure imbalance.
|
On February 13 2014 06:41 imrusty269 wrote:Show nested quote +On February 13 2014 05:59 DinoMight wrote:On February 13 2014 05:40 SC2Toastie wrote:On February 13 2014 05:29 DinoMight wrote:On February 13 2014 05:13 Destructicon wrote:On February 13 2014 04:56 DinoMight wrote:On February 13 2014 04:43 Destructicon wrote:
Well, due to imbalance what turns out happening in tournaments is this.
A lot of weaker representatives of their race qualify for a tournament due to their race being stronger at a time.
That's sign number one of imbalance, over representation in qualifiers if all else is equal.
Second thing that happens is that, in the lower parts of the main tournaments the masses of bad players that where carried by imbalance get slowly culled by the way better players of the other race, giving the false perception of balance by skewing win rates closer to 50%. This is another sign of imbalance but a much more subtle one.
Lastly when players of equal skill meet up in the higher up parts of the tournament then the imbalanced race starts dominating again and they occupy most of the podium spots. This is the last sign of imbalance.
You can observe all these signs of imbalance in all the previous periods of massive domination from one race or another. So i guess you could look at like... race ratio in first round / qualifier vs. race ratio in Ro16/8 vs. podium finishes. I don't know if there is enough data for something like this to be statistically significant though (in the academic definition of significant). You'd need a ton of tournaments and players. You're right though, tournament win rates are kind of silly to look at for balance. Because Tajea is going to beat a mediocre Protoss regardless of whether or not the game is balanced. Taeja can even beat some top tier Protosses despite any imbalance that may exist, because he's really really good. I think they should just take (non mirror wins for race x) / (non mirror games played by race x) and literally just do this for every single ladder game played total. You can't do it by league because the leagues are skewing the data to 50% intentionally. Wouldn't that technically show balance or imbalance>? Actually no, about 4-5 tournaments over a period of 3 months is enough data usually. I'd say IEM Sao Paulo, ASUS RoG, and the Code B and Code A pooled data is nearly enough, Code S finishing would be even more conclusive evidence. WCS EU is kind of poor because we have a huge mix of different skill levels because Koreans out match most of the EU players. And pooling all ladder data is useless since you'll see tons of low level plays. Data should only be collected from Masters and GM, Korean qualifiers or any qualifiers with lots of Koreans involved and premier tournaments with a lot of Koreans. Edit: And again, by data I mean a combination of win rates and race distribution across ladder and tournaments. But if you're only selecting Masters and GM then you run into the same problem you were just discussing where the ladder forces a 50% winrate regardless of how good you actually are. How many people in GM actually have a win rate statistically > 50%?? I also don't think you should be looking at only Koreans. So I think maybe if you looked at winrate averages for those players who are good enough to be statistically above the 50% that Blizzard forces you to you could get an idea of racial balance. If all of them are Protoss or Terran or Zerg then you'd know that the only way to be better than everyone else is to play this race. Well, you wouldn't know for sure but it would be likely. Racial distribution in masters/GM says something. Also, you can look at foreign pros, but they are all relatively 'bad' and will all lose to korean pros (except Scarlett, Naniwa and a select few), and thus scew numbers. Well, I think Protoss has the most to gain from practicing on ladder as they're the agressors early in the two non mirror matchups. So it makes sense that they would play a lot more ladder and therefore rank higher / in more numbers than Terran/Zerg. A lot of really really good Korean Terrans are nowhere to be found on the ladder.. From what I've heard a lot of the Terran/Zerg players practice in customs and don't play so much ladder. It's easy to say Protoss imba because obviously they have the most players in Masters/GM but I think there are many other factors that play into this.. honestly I think global win rates would be, statistically, the best test. Sure it would be messy as hell, but all things equal, if you took every game of starcraft played (ladder AND custom) and did vanilla win rates, I see no reason for one race to have more win % than another unless there is imbalance. Does vanilla wn rate include bronze level match ? How is that useful. If you only include from certain levels up then run into same proplem. First paragraph such trolling lol. You try so hard to defend protoss it makes you look silly.
I really wish this forum had a multi quote.... -____-;
I'm not trolling. And I don't know if you saw any of my posts or actually read what I'm trying to argue, but I'm not tryin to say any race is imbalanced or not. I'm just giving suggestions of ways to calculate "balance." Because there are a lot of factors that go into balance.
You on the other hand came in, saw my Protoss avatar, and started flaming. Read a few posts up and there is actually an educated conversation going on about proper ways to measure things statistically and factors that impact "winrates" that you just interrupted.
EDIT - and yes you should count Bronze matches. Because if a Bronze Protoss consistently beats a Bronze Terran, then that could be an indication of imbalance. Sure Bronze Protosses will lose to Masters Terrans, but also Bronze Terrans will lose to Masters Protoss. It's all about percentages and if the deviation is statistically significant.
|
On February 13 2014 06:36 Destructicon wrote:Show nested quote +On February 13 2014 05:59 DinoMight wrote:On February 13 2014 05:40 SC2Toastie wrote:On February 13 2014 05:29 DinoMight wrote:On February 13 2014 05:13 Destructicon wrote:On February 13 2014 04:56 DinoMight wrote:On February 13 2014 04:43 Destructicon wrote:
Well, due to imbalance what turns out happening in tournaments is this.
A lot of weaker representatives of their race qualify for a tournament due to their race being stronger at a time.
That's sign number one of imbalance, over representation in qualifiers if all else is equal.
Second thing that happens is that, in the lower parts of the main tournaments the masses of bad players that where carried by imbalance get slowly culled by the way better players of the other race, giving the false perception of balance by skewing win rates closer to 50%. This is another sign of imbalance but a much more subtle one.
Lastly when players of equal skill meet up in the higher up parts of the tournament then the imbalanced race starts dominating again and they occupy most of the podium spots. This is the last sign of imbalance.
You can observe all these signs of imbalance in all the previous periods of massive domination from one race or another. So i guess you could look at like... race ratio in first round / qualifier vs. race ratio in Ro16/8 vs. podium finishes. I don't know if there is enough data for something like this to be statistically significant though (in the academic definition of significant). You'd need a ton of tournaments and players. You're right though, tournament win rates are kind of silly to look at for balance. Because Tajea is going to beat a mediocre Protoss regardless of whether or not the game is balanced. Taeja can even beat some top tier Protosses despite any imbalance that may exist, because he's really really good. I think they should just take (non mirror wins for race x) / (non mirror games played by race x) and literally just do this for every single ladder game played total. You can't do it by league because the leagues are skewing the data to 50% intentionally. Wouldn't that technically show balance or imbalance>? Actually no, about 4-5 tournaments over a period of 3 months is enough data usually. I'd say IEM Sao Paulo, ASUS RoG, and the Code B and Code A pooled data is nearly enough, Code S finishing would be even more conclusive evidence. WCS EU is kind of poor because we have a huge mix of different skill levels because Koreans out match most of the EU players. And pooling all ladder data is useless since you'll see tons of low level plays. Data should only be collected from Masters and GM, Korean qualifiers or any qualifiers with lots of Koreans involved and premier tournaments with a lot of Koreans. Edit: And again, by data I mean a combination of win rates and race distribution across ladder and tournaments. But if you're only selecting Masters and GM then you run into the same problem you were just discussing where the ladder forces a 50% winrate regardless of how good you actually are. How many people in GM actually have a win rate statistically > 50%?? I also don't think you should be looking at only Koreans. So I think maybe if you looked at winrate averages for those players who are good enough to be statistically above the 50% that Blizzard forces you to you could get an idea of racial balance. If all of them are Protoss or Terran or Zerg then you'd know that the only way to be better than everyone else is to play this race. Well, you wouldn't know for sure but it would be likely. Racial distribution in masters/GM says something. Also, you can look at foreign pros, but they are all relatively 'bad' and will all lose to korean pros (except Scarlett, Naniwa and a select few), and thus scew numbers. Well, I think Protoss has the most to gain from practicing on ladder as they're the agressors early in the two non mirror matchups. So it makes sense that they would play a lot more ladder and therefore rank higher / in more numbers than Terran/Zerg. A lot of really really good Korean Terrans are nowhere to be found on the ladder.. From what I've heard a lot of the Terran/Zerg players practice in customs and don't play so much ladder. It's easy to say Protoss imba because obviously they have the most players in Masters/GM but I think there are many other factors that play into this.. honestly I think global win rates would be, statistically, the best test. Sure it would be messy as hell, but all things equal, if you took every game of starcraft played (ladder AND custom) and did vanilla win rates, I see no reason for one race to have more win % than another unless there is imbalance. I'm quite sure that during the BL Infestor era the ladder in masters and GM was still dominated mostly by zergs, just like its now dominated by protoss in terms of racial distribution.
I was just thinking about that. I would like to see those statistics as well. If we were to find that even back then the ladder was all Protoss it would be quite signifiant IMO.
|
On February 13 2014 06:41 DinoMight wrote:
I'm precisely trying to avoid to jump to connections of causality, but one theory could be that as Protoss if you want to practice something you can just go on ladder and practice it. But as Terran and Zerg you could be practicing something and then get Oracle'd. Or Blink allined, or 4 gated.. etc. Protoss dictates the game in those early stages a lot. So ladder is less effective to practice something and therefore you wouldn't play as much on it.
I get what you're saying; but in this perspective, ladder is a good simulation of a Protoss who varies his play a lot, since you will meet everything, from all-ins to quick thirds including various cheeses.
The statements I made are not fact, just an opinion. I saw in interviews innovation and MMA say they play a lot of customs w/ Scarlett, but Dear said he mostly just plays ladder. Some other pros made similar comments in interviews. That's what I was going off of.
OK, but you cannot infer their custom/ladder ratios just from interviews. Bogus has a ladder account with quite a lot of games on the KR server, for instance (one of the two top16 GM Terrans with some tag is him); maybe he just plays extra customs too, or he was preparing a specific match, etc.
I don't know how Koreans trains but on the Europe server, I'm sure overall ladder outweights customs by far for all races.
|
4713 Posts
On February 13 2014 06:59 TheDwf wrote:Show nested quote +On February 13 2014 06:41 DinoMight wrote:
I'm precisely trying to avoid to jump to connections of causality, but one theory could be that as Protoss if you want to practice something you can just go on ladder and practice it. But as Terran and Zerg you could be practicing something and then get Oracle'd. Or Blink allined, or 4 gated.. etc. Protoss dictates the game in those early stages a lot. So ladder is less effective to practice something and therefore you wouldn't play as much on it. I get what you're saying; but in this perspective, ladder is a good simulation of a Protoss who varies his play a lot, since you will meet everything, from all-ins to quick thirds including various cheeses. Show nested quote +The statements I made are not fact, just an opinion. I saw in interviews innovation and MMA say they play a lot of customs w/ Scarlett, but Dear said he mostly just plays ladder. Some other pros made similar comments in interviews. That's what I was going off of. OK, but you cannot infer their custom/ladder ratios just from interviews. Bogus has a ladder account with quite a lot of games on the KR server, for instance (one of the two top16 GM Terrans with some tag is him); maybe he just plays extra customs too, or he was preparing a specific match, etc. I don't know how Koreans trains but on the Europe server, I'm sure overall ladder outweights customs by far for all races.
I won't pretend I'm an authority on Korean practice methods, but TB mentioned a couple of times that now a days, the main practice method for Koreans is ladder. Koreans use ladder to train, sharpen and hone their mechanics and fundamentals, and to maintain those at a high level. However when they are working on new build orders or preparing snipe builds they usually train in house via customs.
I guess in Proleague, the best practice method is probably just grinding a specific strategy on a specific map over and over again.
However for weekend tournaments the fundamentals are more important.
|
On February 13 2014 07:32 Destructicon wrote:Show nested quote +On February 13 2014 06:59 TheDwf wrote:On February 13 2014 06:41 DinoMight wrote:
I'm precisely trying to avoid to jump to connections of causality, but one theory could be that as Protoss if you want to practice something you can just go on ladder and practice it. But as Terran and Zerg you could be practicing something and then get Oracle'd. Or Blink allined, or 4 gated.. etc. Protoss dictates the game in those early stages a lot. So ladder is less effective to practice something and therefore you wouldn't play as much on it. I get what you're saying; but in this perspective, ladder is a good simulation of a Protoss who varies his play a lot, since you will meet everything, from all-ins to quick thirds including various cheeses. The statements I made are not fact, just an opinion. I saw in interviews innovation and MMA say they play a lot of customs w/ Scarlett, but Dear said he mostly just plays ladder. Some other pros made similar comments in interviews. That's what I was going off of. OK, but you cannot infer their custom/ladder ratios just from interviews. Bogus has a ladder account with quite a lot of games on the KR server, for instance (one of the two top16 GM Terrans with some tag is him); maybe he just plays extra customs too, or he was preparing a specific match, etc. I don't know how Koreans trains but on the Europe server, I'm sure overall ladder outweights customs by far for all races. I won't pretend I'm an authority on Korean practice methods, but TB mentioned a couple of times that now a days, the main practice method for Koreans is ladder. Koreans use ladder to train, sharpen and hone their mechanics and fundamentals, and to maintain those at a high level. However when they are working on new build orders or preparing snipe builds they usually train in house via customs. I guess in Proleague, the best practice method is probably just grinding a specific strategy on a specific map over and over again. However for weekend tournaments the fundamentals are more important.
IMO:
Ladder is just basic practice, rehearsal, everything can happen and you are forced to play on you toes the entire time and sharpen your mechanics.
Customs are for when you want to practice something specific, a certain build, a certain situation in a game, etc.
I agree with Destructicon on the tournaments.
|
Not surprisingly, they all look quite similar.
![[image loading]](http://i.imgur.com/imx4olG.png)
![[image loading]](http://i.imgur.com/3f0OxtB.png)
![[image loading]](http://i.imgur.com/ZVGG3iQ.png)
|
On February 13 2014 00:35 Ghanburighan wrote:
Btw, TheDwf, I updated my post with Jjakji's actual contribution in terms of win percentages while you were typing.
The question is why you do not count out the games of the best (results) protoss as well. Altering statistics in your favour seems to be one of your specialities.
The low number of terran games in the current context is evidence enaugh that balance is not well at the moment tho.
Still this changes nothing about my first sentence. You just always come up with heavily biased views and altered statistics whenever I read any of your posts.
I also wonder why you didn't complain when like 11 terran qualified first for last year's blizzcon while there was barely any other race's player who qualified for it at the same time.
Also you seem not to know that SC2 was heavily dominated by terrans over the longest periods of time besides the time of BL/Infestor and are still longing for pre queen buff times. You not even seem to worry about it but instead you obviously want to get these times back where terrans got freewins in 8 minute games with one out of 5 random all-ins in TvZ e.g. that were hardly holdable at all, even if scouted, etc. You are just so biased that your posts have actually no value for this discussion at all. You even wrote terran up posts while it was clearly not. You were one of these guys who argued that it is just flash and innovation who only can play that well and counted out there stats in last years statistics the same way you count out the other guy now (all that while taeja, maru, bomber ... had their best year). I wonder where terran advocats take the right from to count out their best guys in statistics no matter if terran is currently op or up just in order to support their biased points of view and advocats of other actually races don't and never do.
According to your last years statements you should now count out the two best protoss players and alter the statistics into the other direction as this is what you did last year with innovation & flash. But you rather turn things around to your needs. When life lost a game to sjow it was life being bad. When innovation/flash (or any other terran) lost a game vs a foreinger it was a clear sign of imbalance. This is the definition of bias. And you are the definition of bias amongst a few others in here, where I am not even sure if you are not one of their alts. This is getting boring and has always been obvious mate. :|
|
Not going to dignify that with a response.
|
You don't need to dignify anything. It is a clear proof of bias to count out last year innovation/flash from your statistics in order to proof T is not op and now count out jjakji in order to proof P is even more op. The actual state of balance is not even important for detecting this. You just take arguments for the race that you play yourself and don't let them count the other way round later on. It is even more, you use the same arguments that you used before but the other way round now. :x
If you were an honest guy you would do neither of these. And if at all do it next time both the same way (e.g. count out the 2 most successful players of the race that is estimated to be superior).
What I came up with here is a clear proof of you being biased all through (q.e.d.) and I hardly doubt anyone can deny this, no matter what the current/past state of balance in fact is/was. Get along with it.
|
|
|
|
|
|
EPT
