|
Finland362 Posts
Update 13 May 2011: added patch 1.3 stats.
I have been keeping track of pro-level games going on for a while with the help of TLPD. I've also added a lot of the missing games myself and organized the games by patch to see how the balance of SC2 has evolved after release. Because I like numbers, I also made up a name for one that describes the overall balance of a patch or map: the Inherent Matchup(/Map) Balance Assessment - IMBA. It's basically the sum of how much the win rate in each matchup (TvZ, ZvP, PvT) differs from 50%. For example if the win rates for those matchups are 60%, 50% and 55% respectively, the IMBA is 10%+0%+5% = 15%.
Since 1.2, the data is now just games that appear on the TLPD. Note that the early patches have quite a small sample size in some matchups, so those shouldn't be taken too seriously. From 1.2 onwards, the margin for error for each matchup is around +/- 3%.
Patch 1.0 - 1.0.3
Non-mirror games: 967
IMBA: 19.5%
TvZ: 180-120 (60.0%)
ZvP: 77-109 (41.4%)
PvT: 236-245 (49.1%)
Patch 1.1 - 1.1.1
Non-mirror games: 406
IMBA: 20.6%
TvZ: 65-45 (59.1%)
ZvP: 24-31 (43.6%)
PvT: 108-133 (44.8%)
Patch 1.1.2 - 1.1.3
Non-mirror games: 1689
IMBA: 7.0%
TvZ: 330-306 (51.9%)
ZvP: 189-207 (47.7%)
PvT: 310-347 (47.2%)
Patch 1.2.0 - 1.2.2 - Now with all TLPD games!
Non-mirror games: 4002
IMBA: 8.3%
TvZ: 692-556 (55.4%)
ZvP: 534-573 (48.2%)
PvT: 842-805 (51.1%)
Patch 1.3 - 1.3.2
Non-mirror games: 3358
IMBA: 6.9%
TvZ: 619-554 (52.8%)
ZvP: 516-461 (52.8%)
PvT: 588-620 (48.7%)
Graph of how the matchups have played out in each patch:
![[image loading]](http://i.imgur.com/UdIRt.jpg)
+ Show Spoiler [Out of date map stats] +
|
I really like the matchup stats, but you really can't analyze the maps without separating by patch. Kulas was an imbalanced map, yes, but it's made a lot worse in these stats by the fact that it was only played under a balance condition that was pretty bad regardless of the map. Shakuras has had much more of its play under more balanced patches.
|
Close position metal is as bad if not worse than steppes. Zerg has a much harder time taking a third.The only other big map is Shakuras which is the most balanced because positions don't offer as great of a variation
|
Shakuras Plateau seems to be really balanced, hopefully I will get to play it more on the ladder...oh wait....nvm.
|
On January 19 2011 22:25 Dont Panic wrote:
Close position metal is as bad if not worse than steppes. Zerg has a much harder time taking a third.The only other big map is Shakuras which is the most balanced because positions don't offer as great of a variation
thats so very true, on steppes at least you are in an uphill position while terrans hold the hill and are able to crawl into your natural
i personally find shakuras is the most boring map where games usually end up into a macro festival where none attacks within the first 15mins aside from a little harassment attempts
im also not surprised that there are so many games on metalopolis as in my own case almost half my games are exactly there
|
Wow, nice stats. They just prove what I was claiming all along from the start of SC2 to people on another forum, that Terrans dominate all tournaments. The ONLY surprising thing here is the TvZ on Steppes of War.
Anyways, I will eagerly await stats for maps based on recent patches only.
|
TvZ in steppes is pretty hard sometimes because you get 0 practice on that map against zerg if you play ladder and against a zerg who trained the map a lot he might have an ace up his sleeve.
Not saying the map is anything but horrible but the wins i've seen from zerg has been terrans losing due to doing outdated or poor things.
Shakuras is my favorite map. I do feel that it should be three stones blocking (or two with greater health) but other then that I think it's a great map and it blows my mind it's gone from the ladder
|
Why would we rage at Metalopolis looking so bad?
Fact of the matter is that it's an awful map unless it's cross positions.
|
Imo on maps like metalopolis, it should be codeded into the game that players start cross-position from one another.
|
On January 19 2011 23:29 phexac wrote:
Imo on maps like metalopolis, it should be codeded into the game that players start cross-position from one another.
As a biased zerg player, I agree 200%. IMO it's really frustrating that so much luck comes in to play with spawn positions... not that you can't win close positions but I guarantee if we had the stats on close vs. cross positions it would be scary.
|
Glad there's some stats now to back up the people saying Meta is bad for zerg. So tired of people saying otherwise- namely Tastosis raving about what a good map it is.
Two positions are as close to auto-loss as you can get. Close positions is way worse than Steppes. Close air positions and you have no angle of attack except through a narrow low-ground corridor of death.
|
On January 20 2011 01:43 Neo.NEt wrote:Show nested quote +On January 19 2011 23:29 phexac wrote:
Imo on maps like metalopolis, it should be codeded into the game that players start cross-position from one another. As a biased zerg player, I agree 200%. IMO it's really frustrating that so much luck comes in to play with spawn positions... not that you can't win close positions but I guarantee if we had the stats on close vs. cross positions it would be scary. Ask and you shall receive(original thread posted by heyoka).
|
Combined win percentages for each race- possibly useful in choosing which maps to thumbs down.
Zerg
Shakuras Plateau 102
Steppes of War 95
Scrap Station 96
Blistering Sands 96
Lost Temple 91
Xel'Naga Caverns 90
Delta Quadrant 91
Jungle Basin 80
Metalopolis 78
Desert Oasis 58
Kulas Ravine 69
Terran
Shakuras Plateau 100
Steppes of War 103
Scrap Station 109
Blistering Sands 103
Lost Temple 110
Xel'Naga Caverns 105
Delta Quadrant 117
Jungle Basin 111
Metalopolis 112
Desert Oasis 105
Kulas Ravine 129
Protoss
Shakuras Plateau 98
Steppes of War 102
Scrap Station 95
Blistering Sands 101
Lost Temple 99
Xel'Naga Caverns 105
Delta Quadrant 92
Jungle Basin 107
Metalopolis 110
Desert Oasis 137
Kulas Ravine 102
|
metalopolis is great for zerg if its non-close pos. if its close pos its very bad for zerg :p
nice statistics ! 
i still think xel naga, metal (non close pos) and shakuras are the best maps. but not because of balance, but because of gamestyle and that 90% of the games doesnt turn into cheese
|
Wow great information OP. This is just as I suspected - Zerg really needs a boost, especially on certain situations where maps give them unfairly high losing percentages.
|
EDIT: oh dam posted in a balance discussion thread and notice only 3 sek after i clicked post.
Whats wrong with me?
|
I hope we see larger maps soon, and fixing metal start positions like shakuras as well as re-adding shakuras would really be nice. I can't stand the current map pool. No one likes watching a game boot up with that sick feeling in your stomach of plz don't spawn me an auto-loss.
|
Great statistics, very useful. Seems ridiculous how two maps are almost 50% unbalanced.
|
Somewhat surprising to see the matchup balance for Steppes to be so high, haha. Maybe the fact that you already know that your opponent is spawning very close to you has to do with it (whereas in Metalopolis blindly going for early gas early banelings may screw you over hard).
|
On January 20 2011 02:05 Alsn wrote:Show nested quote +On January 20 2011 01:43 Neo.NEt wrote:On January 19 2011 23:29 phexac wrote:
Imo on maps like metalopolis, it should be codeded into the game that players start cross-position from one another. As a biased zerg player, I agree 200%. IMO it's really frustrating that so much luck comes in to play with spawn positions... not that you can't win close positions but I guarantee if we had the stats on close vs. cross positions it would be scary. Ask and you shall receive(original thread posted by heyoka).
Awesome, you're the man.
|
That was really cool OP for you to actually spend the time and put that up here though.
|
On January 19 2011 22:25 Dont Panic wrote:
Close position metal is as bad if not worse than steppes. Zerg has a much harder time taking a third.The only other big map is Shakuras which is the most balanced because positions don't offer as great of a variation
This is exactly it. In fact, its not only the difficult to take 3rd base. But its also due to the high ground expansion between the bases in proxy positions.
While Steps is such a short rush, if you don't die with in the first 5 mins, the map actually is decent come Mid / late game.
|
Wait, how does this proves that Z is UP/needs a boost?
The latest patch percentages has all matchups around 50%, unless I am missing something >_>
Anyways, yeah Close positions Metal is pretty bad for Zerg, at least in steppes the natural is easier to defend.
|
Great job man! Really great, I find that the ONLY source for balance should be pro level gamers.
In addition to that it corresponds to my feeling as a protoss, weaker vs T than vs Z!
|
|
|
Some of the numbers are quite small (like...how many matches took place on Kulas?). Including standard errors would be good, so we can see if the IMBA is even statistically significant.
|
On January 20 2011 06:51 Dragar wrote:
Some of the numbers are quite small (like...how many matches took place on Kulas?). Including standard errors would be good, so we can see if the IMBA is even statistically significant.
So 100,000 match proves something but 100-10 doesn't? If one needs this many games to get the point, one will not get the point anyway.
|
On January 19 2011 22:25 Dont Panic wrote:
Close position metal is as bad if not worse than steppes. Zerg has a much harder time taking a third.The only other big map is Shakuras which is the most balanced because positions don't offer as great of a variation
i cannot remember the last time I won a close position meta as zerg against terran.
it seems i play meta like 50% of my games, and i get close positions 75% of those games. its so difficult to win against, what the hell do you do when they get a PF at the third and their tanks defending the third can shoot ur nat?
i wish i could thumbs down 6 maps in this game
As a biased zerg player, I agree 200%. IMO it's really frustrating that so much luck comes in to play with spawn positions... not that you can't win close positions but I guarantee if we had the stats on close vs. cross positions it would be scary.
im really glad people feel my pain, its def my most played map too. honestly when i play on meta i just have to cross my fingers and hope i get good LUCK
|
This is a good study, thanks for analysing the games so far. I hope there are people like you working as Blizzard balance squad.
|
good job. as always people will point out the flaws of this but it probably took a lot of hard work and it always gives people a sense of right direction and thats how it is in science.
|
Finland362 Posts
Now that version 1.3 is released I think it's time to bump this with an update again. As the number of tournaments really got out of hand, I wrote some Python scripts to automatically parse all TLPD games. Here are the new stats:
Patch 1.2.0 - 1.2.2 - Now with all TLPD games!
Non-mirror games: 4002
IMBA: 8.3%
TvZ: 692-556 (55.4%)
ZvP: 534-573 (48.2%)
PvT: 842-805 (51.1%)
The changes in 1.2.0 weren't exactly sweeping so most likely the changes from the previous version are from map pool changes and evolution of strategies. As for maps, the amount of games played on non-Blizzard maps is still too low to say anything about them.
|
You rock. Hope you keep this up after 1.3.
|
|
|
Wow you rock. Hopefully we can keep this up
|
Really interesting to see that the race vs race stats are still pretty similar as they were before as to which races win the most and which lose the most etc. No real surprises though, and still pretty cool to see some real stats.
|
Very interesting, thanks!
|
Finland362 Posts
Here come the stats for 1.3 through 1.3.2.
Patch 1.3 - 1.3.2
Non-mirror games: 3358
IMBA: 6.9%
TvZ: 619-554 (52.8%)
ZvP: 516-461 (52.8%)
PvT: 588-620 (48.7%)
To recap: major changes from 1.2 were increasing stim research time, removal of Khaydarin Amulet and the buff to Fungal Growth.
Now I'll leave you with this pretty picture of the matchup win rates post-release:
|
From the looks of the graph, seems like zergs are winning more games. nice
|
The game seems to be very balanced currently. Race have almost no affect on the outcome in most games, maps maybe a little bit more.
|
Finland362 Posts
Yes, it should be emphasized that all of these numbers are over the map pool that was being played in that patch and it's very hard to separate what's caused by better maps and what's caused by gameplay changes.
|
I wonder what the stats would show if you removed Naniwa from them.. Since he is the only protoss really winning everything lately. Even MC is down significantly.
|
On May 13 2011 19:29 Skyze wrote:
I wonder what the stats would show if you removed Naniwa from them.. Since he is the only protoss really winning everything lately. Even MC is down significantly.
Kiwikaki is making finals but not winning them (MLG IPL), Naniwa had to beat Hasuobs to make TSL finals so there would be a protoss there either way, Inca in GSL finals, MC and Whitera at the dreamhack invite. Protoss is doing fine even without Naniwa
|
Current stats look pretty good
TvZ: 619-554 (52.8%)
ZvP: 516-461 (52.8%)
PvT: 588-620 (48.7%)
|
|
|
|
|
|
EPT
