|
GSL's May Tournament, the "Super Tournament" featured 29 Terran players, 19 Zerg players and 16 Protoss Players (64 total) split into 4 brackets. This tournament combined Code A and Code S and brought back some old GSL favorites. With the Round of 64 ending, it is interesting to look at the racial distribution of the Round of 32 and the implications to future rounds.
Round of 64 - 29 Terrans, 19 Zergs, 16 Protoss
Code S vs Code A/B Results (provided by Tachion)
23 Code S vs Code A/B Games
Code S 15 wins - ~65% win ratio
Code A/B 8 wins - ~35% win ratio
Non-Mirror Match Set Statistics
TvP 6 Sets - 5 to 1 - ~83% of Terrans Advance, ~17% of Protoss
TvZ 7 Sets - 4 to 3 - ~57% of Terrans Advance, ~43% of Zergs
PvZ 6 Sets - 4 to 2 - ~ 66% of Protoss Advance, ~33% of Zergs
Non-Mirror Match Game Statistics
TvP 15 Games - 11 to 4 - ~73% win ratio for Terran, ~26% for Protoss
TvZ 16 Games - 8 to 8 - 50% win ratio for Terran, 50% for Zerg
PvZ 16 Games - 9 to 7 - ~56% win ratio for Protoss, ~44% of Zerg
Total Non-Mirror Match Set and Game Win Percentages
Terran - Set ~69% Game ~61%
Zerg - Set ~38% Game ~47%
Protoss- Set ~42% Game ~42%
Total Number of Players Advancing from Non-Mirror Matches Per Race
9 of 13 Terrans Advance - ~69%
5 of 13 Zergs Advance - ~38%
5 of 12 Protoss Advance - ~42%
Number of Mirror Matches Per Race
Terran - 8
Zerg - 3
Protoss - 2
Overall Results
~58% of Terrans Advance
~42% of Zerg Advance
~43% of Protoss Advance
Round of 32 - 17 Terrans, 8 Zergs, 7 Protoss
Round of 16 - 4 Terrans and 1 Zerg in the Ro16 based on Ro64 Results
Round of 8 - 1 Terran in the Ro8 based on Ro64 Results
|
PvZ stats look the wrong way round?
Nothing unexpected. Way too many TvTs 
|
very small samples... but there is a terran player dominance and that may or may not be linked to game balance.
|
No wonder TvZ is always the most fun matchup to watch, it seems to be the most balanced one in terms of competition.
|
|
|
ye small sample size but this is what it looks like in pretty much all tournaments. There allot more terrans than zerg/protoss (has nothing to do with balance) and therefor so many TvT's.
I'm shocked at the 78% win rate TvP...and terrans are seriously QQing over Infestor/zerg in general lol.
|
It is interesting that so many Terrans were in the Super Tournament compared to Zerg or Protoss players. You'd think Gom would try and even it out more. Are there really that many good Terrans compared to Zerg or Protoss? If yes, what does it say?
However, it is when you remove the mirror matches from the non-mirror matches that the results get really interesting. TvZ was by far the most balanced, while PvZ favored the Protoss. TvP heavily favored the Terrans.
|
On June 01 2011 00:42 BronzeKnee wrote:
Round of 16 - 4 Terrans and 1 Zerg in the Ro16 based on Ro64 Results
Round of 8 - 1 Terran in the Ro8 based on Ro64 Results
This looks quite strange. :D
Btw, interesting stats. Thank you. 
|
Sorry but too small sample to make this even remotely interesting.
|
Terran is doing really well in Korea according to SC2Ranks. 9 of the top 10 spots belong to Terran players.
I'm not really sure I understand why, though...
|
On June 01 2011 01:20 Longshank wrote:
Sorry but too small sample to make this even remotely interesting.
The small sample size is limiting, but how many good players in the world are there that can play at this level? We see some players we thought were "good" (Painuser, QXC) crashing and burning in the NASL.
So if we increase the sample size too much it ruins the validity of the results too, since you'll just continue increasing the gigantic variable that is player skill.
|
To the people say small sample size not relevant:
this is a population not a sample, since this is composed of the entirety of the super tournament.
Take a quick refresher if you think this is a sample size.
Edit: I think that the predictions are based on facts due to the bracket distribution, but I could be wrong on this.
|
It doesn't change the fact that you shouldn't try to analyze too much from a tiny sample. There's just too much noise.
|
On June 01 2011 01:23 MrBitter wrote:
Terran is doing really well in Korea according to SC2Ranks. 9 of the top 10 spots belong to Terran players.
I'm not really sure I understand why, though...
1. All ins are much stronger and impossible to scout since they can always just "fake" them therefor until ovy peed/obs you can't be cure
2. APM/control/awareness of what your opponent is doing needed to play is much lower thus on ladder where ppl don't play there best a race that requires 50 apm is superior to one that requires 150 to be played properly
3. The mirror match up isn't coin flip like the other 2 therefor a terran pro will have about 70-80% WLR on ladder in all match up while a Z/P one will have 70-80% in 2 of them and around 50-60% in mirror
4. The race is freaking overpowerd ? since the game begun most pros agreed that terran > toss > zerg, only that for about 2 months ppl forgot that since they were to busy QQing about PvZ until zerg discovered drops, banelings, queens, roaches and ultras, now that it is over ppl will go back to complaining about terrans im quite sure.
|
On June 01 2011 01:30 slyboogie wrote:
It doesn't change the fact that you shouldn't try to analyze too much from a tiny sample. There's just too much noise.
This is correct, but there are plenty of issues with a large sample size too.
In general we want to control variables as much as possible. Or as my quanitative methods teacher explained, "there is no free lunch!" Meaning in this case, that we can infer that these results are accurate for the GSL (to an extent, since player skill is quite a variable) and perhaps other high end tournaments, but perhaps not SC2 as a whole. If we increase the sample size too much though, the results are no longer apply to the GSL or other high end tournaments, and then perhaps even the results are indicative of nothing at all.
Thus, we cannot gain information for free with a greater sample size since we are also introducing more variables. Looking at the GSL alone is like looking at a small piece of large picture, while looking at a large sample size is similar to looking at a large picture from far away. Neither gives you a complete and clear view, but both are useful.
Also it appears the TSL3 results are similar so far, giving us another chunk of the large picture to look at in clarity.
|
You can't extract any meaningful conclusions from this data. I could just as easily say the better player won almost every game.
|
On June 01 2011 01:37 hmunkey wrote:
You can't extract any meaningful conclusions from this data. I could just as easily say the better player won almost every game.
What you say highlights the normal response to statistics that someone who doesn't want them to say anything says. You picked out an uncontrollable variable and said that all the results are because of said variable.
Certainly player skill is an important uncontrolled variable, which is also why you don't want to increase the sample size too much because you'll increase this variable. Player skill is a relatively well controlled variable in the GSL because you had to qualify through a grueling tournament to get in (unlike the NASL which has a massive difference in player skill). However, when you do begin to increase sample size (add in TSL3 and GSL May) and you'll see the trend (although GSL May Code S was very balanced, Code A was quite the opposite).
There really is no free lunch in statistics, but they never say nothing.
|
Canada13427 Posts
On June 01 2011 01:38 BronzeKnee wrote:Show nested quote +On June 01 2011 01:37 hmunkey wrote:
You can't extract any meaningful conclusions from this data. I could just as easily say the better player won almost every game. What you say highlights the normal response to statistics that someone who doesn't want them to say anything says. You picked out an uncontrollable variable and said that all the results are because of said variable. Certainly player skill is an important uncontrolled variable, which is also why you don't want to increase the sample size too much because you'll increase this variable. Player skill is a relatively well controlled variable in the GSL because you had to qualify through a grueling tournament to get in (unlike the NASL which has a massive difference in player skill). However, when you do begin to increase sample size (add in TSL3 and GSL May) and you'll see the trend (although GSL May Code S was very balanced, Code A was quite the opposite). There really is no free lunch in statistics, but they never say nothing.
No but to be honest some of the players in the super tournament are in thanks to points from earlier seasons even though they have been performing poorly recently.
Also so many Terrans >.< Dang TvTs ahead lol
|
On June 01 2011 01:51 ZeromuS wrote:Show nested quote +On June 01 2011 01:38 BronzeKnee wrote:On June 01 2011 01:37 hmunkey wrote:
You can't extract any meaningful conclusions from this data. I could just as easily say the better player won almost every game. What you say highlights the normal response to statistics that someone who doesn't want them to say anything says. You picked out an uncontrollable variable and said that all the results are because of said variable. Certainly player skill is an important uncontrolled variable, which is also why you don't want to increase the sample size too much because you'll increase this variable. Player skill is a relatively well controlled variable in the GSL because you had to qualify through a grueling tournament to get in (unlike the NASL which has a massive difference in player skill). However, when you do begin to increase sample size (add in TSL3 and GSL May) and you'll see the trend (although GSL May Code S was very balanced, Code A was quite the opposite). There really is no free lunch in statistics, but they never say nothing. No but to be honest some of the players in the super tournament are in thanks to points from earlier seasons even though they have been performing poorly recently. Also so many Terrans >.< Dang TvTs ahead lol
Absolutely true. However, what happens when these poor players play poor players? Someone has to win right? Or what if these poor players all improve drasticially (think San) or previously good players begin performing poorly (Fruitdealer)?
Thus we have Choya advancing, and this absolutely destroys that logic. And then have Inca and MC losing in the first round. So from the Protoss point of view at least, "good" previous performers are going home, while "poor" previous performers are advancing. So are the good performers really good and are the bad performers really bad? If you define good and bad with a definition from the dictionary, the answer is no. Based on the results it is in fact the opposite.
In the end we can't control for player skill in the tournament itself. However, the fact these guys have qualified controls for it to some extent, which is better than no extent (ie super high sample size). Again, it goes back to being able to look clearly at a small piece of a large picture, or be able to see the whole picture from very far away, where you can't make everything out clearly, and certainly can't see any details.
Both are useful, but looking at the small piece of the large picture is almost always what researchers do, since the information is far eaiser to obtain generally.
|
this just further supports my point that korean terran are just better than other koreans and foreigner zerg/toss are better than foreigner terran
|
|
|
|
|
|
EPT
