komi statistics

16 posts / 0 new
Last post
macelee
komi statistics

I just generated some statistics from Go4Go database. As of today, excluding those few games ending with no results, there are

11256 games with 5.5 komi, of which 5961 (52.96%) won by black and 5295 (47.04%) won by white.
24728 games with 6.5 komi, of which 12507 (50.58%) won by black and 12221 (49.42%) won by white.
15010 games with 7.5 komi, of which 7127 (47.48%) won by black and 7883 (52.52%) won by white.

Apparently the smaller 5.5 komi was too easy for black. The increase of komi around year 2003 was a sensible decision. Unfortunately for China, it doesn't make sense to use 6.5 komi under Chinese counting rule. But 7.5 komi clearly gives white a big advantage. Indeed many top Chinese players (e.g. Ke Jie) prefer to use white.

Bloemen
So a komi of 5,5 points seems best to me

I would say that the result of a game should be the same, independent of whether Chinese or Japanese rules are used. It is clear that whatever value for komi is chosen, one side will do better than the other side. It seems natural to me that that value is chosen where chances of winning are most equal, with a faint advantage for Black, who had that advantage to start with without komi. Why overcompensate.

Kind regards,
Paul

kavindisaya
kavindisaya's picture
Reply to macelee

Hi. I'm currently doing a research about go statistics in my college. Could you please provide me a formal source of this information? I will be more than happy to site you in my research. Thanks.

macelee
Updated information

Hi kavindisaya,

The figures are derived from the database hosted with the website. Unfortunately there's no 'formal' source. But Go4Go database is known for its high quality information so these figures should be very authoritative.

I can give you the latest information:

5.5 komi: black 6016 (52.84%), white 5332
6.5 komi: black 12830 (50.58%), white 12536
7.5 kimi: black 7215 (47.50%), white 7975

Generated from Go4Go database on 2016-03-08.

Nickj
Komi should be integer, and probably 7

I never understood why Go players insist on eliminating drawn games. We see people referring to drawn games as "no result" - they usually aren't "no result", a draw is a perfectly legitimate result. Go will never suffer from an excess of drawn games if komi is an integer; chess has thrived in the West for well over a century with a much higher proportion of drawn games than are conceivable in Go.
If we can imagine a game in which neither player makes any mistake, a draw is the logical result. Therefore, it should be a possible result.

trufolo
I think avoiding draw makes

I think avoiding draw makes the game always interesting.
It is an old rule and i like it Smile

Josperus
@Nickj: I think as well, that

@Nickj: I think as well, that a perfect game of both players should end jigo on the board, even if black has the first move.

element
i think komi 6.5 Appropriate

i think komi 6.5 Appropriate

tapsi
I'm also for komi 7

From the statistics seems clear that 5.5 is not enough and 7.5 is too much. If you want to eliminate draws, you can just say that draws are treated as B or W win (which is what 6.5 and 7.5 means basically). But komi itself should be 7 IMO, and not mixed with your draw preferences.

pookpooi
Macelee, do you have the

Macelee, do you have the latest komi stat? I wanna see if how it changes...

pookpooi
Komi stats

Do you have the latest komi statistics? I wanna know if 6.5 and 7.5 get more balanced or deviate.

macelee
Latest komi stat

6.5 komi: black 15972 (50.58%), white 15592
7.5 kimi: black 9025 (47.38%), white 10012

Generated from Go4Go database on 2017-11-24.

pookpooi
Thank you!

Thank you and sorry for double post as I just know (and should know) that all comment have to be wait for approval first. Anyway look like 6.5 komi maintain its balance while 7.5 komi is even worse than the last year.
By the way I divide by myself and get 50.60% for 6.5 Black and 47.41% for 7.5 Black, are your percentage number discount jigo?

macelee
Re: Thank you!

When I calculated the percentage figures, I did include a few extra games with Void result (mostly triple ko kind of games).

pookpooi
I see

Okay I see, thank you.

MagRes
There may be something I am

There may be something I am missing, but doesn't the stronger player always play as white? Are these stats only for equally ranked players? If the games used are not with "equally" ranked players I would expect white to have a larger winning percentage with a fair komi.