EPT
[GSL] 2018 Code S Season 2 - Ro16 Group C - Page 9
| Forum Index > StarCraft 2 Tournaments |
|
DieuCure
France3713 Posts
| ||
|
Boggyb
2855 Posts
On June 01 2018 00:08 Tyrhanius wrote: Maybe people will now finally realize how poorly the zerg race perform at the highest level, but i doubt it. Both ZvT and ZvP are Zerg favored. | ||
|
Tyrhanius
France947 Posts
According to aligulac xd ? Look which matches they add to do their stats : A small sample : Misato PL Z 3–2 T PL bleidd GOLDMADMAN CN Z 2–0 P US BlueApril TY KR T 2–0 P KR Gyulzzing Dark KR Z 2–0 P UA Lisandro Ryung KR T 2–0 Z CN Ganxieni danran CN Z 0–2 Z KR Rogue Solar KR Z 2–0 P KR 퇴근 Impact KR Z 2–0 P CN Firefly Arctur RU T 1–0 Z CH Cammelot Keiras CZ Z 1–0 P UK Floru etc... from such a great tournaments : "Diamonds Are Not Forever Invitational #1", "Kung Fu Cup Weekly 2018 ", " Alpha SC2 Team League Season 11 Euro Pro Group A Houkago Tea Team vs Romandy Gaming Academy", Proxy Tempest EU Open #135 Ro32", etc... and they mix that with GSL, olimoleague, like it's relevant... | ||
|
Boggyb
2855 Posts
Maru is miles above Solar at the moment and Zest mind gamed Solar and hit him with a new build. soO choked hard against Dear. I haven't actually watched soO INnoVation, but ZvT has historically been soO's weakest matchup, so that's not a surprise. Dark played like pure trash against Classic and Trap. Rogue choked hard against Dear. That's a super small sample size and there are a lot better reasons for the results other than imbalance. | ||
|
Athenau
570 Posts
On June 01 2018 05:16 Tyrhanius wrote: According to aligulac xd ? Look which matches they add to do their stats : A small sample : Misato PL Z 3–2 T PL bleidd GOLDMADMAN CN Z 2–0 P US BlueApril TY KR T 2–0 P KR Gyulzzing Dark KR Z 2–0 P UA Lisandro Ryung KR T 2–0 Z CN Ganxieni danran CN Z 0–2 Z KR Rogue Solar KR Z 2–0 P KR 퇴근 Impact KR Z 2–0 P CN Firefly Arctur RU T 1–0 Z CH Cammelot Keiras CZ Z 1–0 P UK Floru etc... from such a great tournaments : "Diamonds Are Not Forever Invitational #1", "Kung Fu Cup Weekly 2018 ", " Alpha SC2 Team League Season 11 Euro Pro Group A Houkago Tea Team vs Romandy Gaming Academy", Proxy Tempest EU Open #135 Ro32", etc... and they mix that with GSL, olimoleague, like it's relevant... Including matches with large skill disparities push winrates towards 50% because the probability of pairing two players with widely differing skill levels is symmetric with regards to race, assuming a similar skill distribution across races. So if Aligulac winrates are close to 50% that doesn't necessarily indicate that the game is balanced because of the noisy data, but for the same reason larger deviations are a good indicator of imbalance because if you stripped out all the mismatched games the deviation would be even larger. | ||
|
Tyrhanius
France947 Posts
On June 01 2018 06:15 Athenau wrote: Including matches with large skill disparities push winrates towards 50% because the probability of pairing two players with widely differing skill levels is symmetric with regards to race, assuming a similar skill distribution across races. So if Aligulac winrates are close to 50% that doesn't necessarily indicate that the game is balanced because of the noisy data, but for the same reason larger deviations are a good indicator of imbalance because if you stripped out all the mismatched games the deviation would be even larger. Which doesn't work, because the region are locked for tournaments. The probabity of a zerg foreign to meet TY on a tournament is really low, while the probablility of a terran foreigners encounters Serral is higher. And there is many more tournaments on EU than on KOR. Some tournaments have BO1 qualifier, some BO3. It really hard to prove here that the number of games is enough to homogenize all the bias. Your hypothesis is the for all games the sum of all difference between MMR of T-Z=DeltaT is only = 0 to 200. I really doubt, it's true, and a fortiori that it was true for all reports. And actually you suppose 45% is different than 50%, which isn't sure. And the problem is it doesn't match the reality, if you see some 65% on TvZ on a tournament while aligulac says 45%, ever aligulac is wrong and fails to show the real state of balance, ever his stat have a confidence interval like 45% +/-20% and it's meaningless to use it. | ||
|
Athenau
570 Posts
Which doesn't work, because the region are locked for tournaments. The probabity of a zerg foreign to meet TY on a tournament is really low, while the probablility of a terran foreigners encounters Serral is higher. Region lock doesn't apply to online tournaments, which is where you'll find the largest skill mismatches. And the same effect holds within a region anyway--assuming similar distributions of skill it's no more likely for a 90th percentile EU Terran to play a 10th percentile EU Zerg than the other way around. Your hypothesis is the for all games the sum of all difference between MMR of T-Z=DeltaT is only = 0 to 200. I really doubt, it's true, and a fortiori that it was true for all reports. No, that's not my hypothesis. And the problem is it doesn't match the reality, if you see some 65% on TvZ on a tournament while aligulac says 45%, ever aligulac is wrong and fails to show the real state of balance, ever his stat have a confidence interval like 45% +/-20% and it's meaningless to use it. Why would you expect the global tournament winrates to be an estimator of an individual tournament's winrates? That's...not how it works. The sample here is tournament games recorded by Aligulac. If anything, you'd expect the Aligulac winrates to be estimators of winrates across all tournament games played. The sample statistic is an estimator of the population parameter not the other way around. | ||
|
juicyjames
United States3815 Posts
| ||
|
sneakyfox
8216 Posts
On June 01 2018 13:50 juicyjames wrote: Anything recommended? Yes, first three series were great. Both Classic-Dark, TY-Trap, and Classic-TY. | ||
| ||
