Komi Go / Discussion

Sub-page of KomiGo

Dieter: It is my belief that komi is not fit for amateurs - say below 5 Dan. If you let me choose colors, I will always take White: 5.5 points in the pocket. We amateurs make so many mistakes that it really doesn't matter who begins. We are certainly not capable of maintaining the advantage of moving first. Professionals are so komi go fits them. The professional level is way above ours. If they decide 8 points is the proper komi I would strongly oppose its adoption in amateur games.

ChessWhiz: Yes, but on the other hand, just as (according to you) we can't maintain a first-move advantage, we can't keep a komi advantage either. I often win and lose games by 10 points or more, and a komi doesn't affect that. So if both players are the same strength, one should start and the other should receive komi. If the pros find that komi should be larger, by all means, make it larger! It simply offsets the advantage of the first move, and I think it's just as hard to retain a komi advantage as it is a first-move advantage. Of course, this is all in my humble opinion. What do the rest of you think?


SAS: Black is half a move ahead (on average) throughout the game; this advantage is there whether you are an amateur or a professional. But I think there is reason to suppose that komi should be smaller for weaker players. My reasoning is as follows. Suppose you and your opponent both drop an average of x points per move (compared to the perfect move). If Black and White both play the same number of moves, then this will cancel out (on average). But in fact Black often plays one move more than White, due to having the first move. If we assume that on average Black plays half a move more than White, then we must conclude that the komi required to make the game fair is x/2 less than the "correct" komi (which is probably 7). This adjustment is very small however - perhaps about 1 point for every 20 stones of strength. So I think a komi of 6 should be about right for most amateurs above 10 kyu. It would be good, however, to see some actual statistics for a few thousand amateur games of about shodan level, to see how things work out in practice.

The point that ChessWhiz makes about the komi not mattering so much in amateur games is correct. The higher x is, the higher its variance is likely to be, and this increases the probability that the winning margin will be large enough that the exact komi (within reason) will not matter.

Lezogzog: There is another point I'd like to make about going first. I like to take black because it allows me to a certain extend to choose my strategy, at least for the first moves. When I play a fuseki I'm more confident with, it makes me more confident for later moves. However, I do agree that it is better not to be too confident. Also, it is sometimes good to play strategies you are not used to, for obvious tutorial reasons. In conclusion, I would say that when I have the choice, I prefer to give komi when I play somebody about my strength (which is btw still low, but who knows).

Tim Brent: I feel for amateurs that 0.5 komi to prevent jigo is sufficient.

SAS: Sufficient for what? And what have you got against jigo?
Tim Brent: I actually have nothing against jigo, and have played one game so far at 0.0 komi. I assume most players want a result (i.e. Black win or White win). That is how I mean, sufficient to assure a result. Another reason for 0.5 to stop jigo is by analogy to Chess, where most games end up drawn, which can be frustrating.
SAS: If all you meant is that 0.5 komi is sufficient to prevent jigo, then you are of course correct (for professionals as well as amateurs). But 6.5 komi is also sufficient for that purpose, and is likely to make Black's winning percentage nearer to 50% (which is the main purpose of komi). Even between professionals, a komi of 7 would probably only cause about 3% of games to end in jigo, so the situation is not comparable to chess.

RobertJasiek: Komi for weaker players is a strategic preparation for playing stronger, so it should already use the stronger players' value. Whether some weaker players notice the strategic effects, depends on every player. E.g., I as a weaker player did notice the effect of even small komi changes and I altered my strategies accordingly.


Graham:

Theoretically
komi should be the value which makes the result a jigo with perfect play for black and white.
Pro games
with komi=5.5 have B winning ~55% and I believe analysis including the W+0.5 results suggests that komi=6.5 is about right. This is also consistent with the results of environmental go where pros value the first move at 13-14 points miai
Practically
far fewer kyu games are settled by komi type score differences, than dan games. This begs the question "Should komi be dependent on rank ?"
Personally
I prefer to take black and would do so up to komi=8. This is principally to have more control of fuseki and belief that komi=5.5 is too low especially for players of my strength (~1kyu). ie W will not have the skills to utilise komi.

Matt Noonan: I've heard this argument that the proper value of komi should make perfect play lead to jigo, but how reasonable is this? You could imagine (maybe for some other game) a proof that black has a strategy which makes him win by at least 25 points without actually giving such a strategy. But would we really start giving white 25 points of komi? I think komi should remain dynamic, the value fiddled with to keep white's winning percentages around 50%. It's kludgy, but it works and it is fair.

Damien Sullivan Would it be reasonable to say that even with komi, even ranked players should be alternating white and black, or at least playing both as frequently, on average? Then small errors in komi will wash out anyway. As said, komi seems a kludge to try to get evenness in a single game, vs. the "natural" balance of playing two games. (Leading to the natural rank of always giving black, and then giving extra beginning moves for handicap.)


What are the suggested komi for non standard board sizes (7x, 9x, 11x, 13x, 15x, 17x, and 21x)? --Hu of KGS

See handicap for smaller board sizes. Charles Matthews


One question (which stems from Damien's note above) is why have komi at all? Yes, black has the advantage of the first move. So what? Why does white need to be "compensated" for a natural consequence of the game? In a match, or even a tournament (generally), you will have, more or less, a chance to play black just as much as white - so you are compensated naturally. It just seems that komi is unnatural and incorrect: the first mover should have an advantage... otherwise, why move first at all?

Alex

Patrick Taylor: I like the fact that komi breaks the game. Supposedly there is a perfect set of black moves against which a perfect set of white moves cannot win. It seems to me--though I could be wrong, I'm not a mathematician--that komi breaks this property. Now, as to the numerical value? It makes sense that the pros set one, because they don't want to disadvantage white in a professional setting. Sure, jubango is probably better, but it would most likely require a comeplete restructuring of professional go. For us? The number could be closely tied to rank. Maybe someone should datamine KGS/IGS to see the percentage of kyu games that are decided by 10 points or less (probably divide SDK from DDK). With this information we might be able to draw some real conclusions.

Velobici: When we get strong enough that none of our mistakes are larger than the komi, I'll start to fret the komi itself. ;)

xela: In this connection, see honinbo#1

Elroch: It would be very interesting to look at the statistics of results at different levels to see if this justified a smaller komi for weaker players. It is possible that the same komi would apply, but that this would be disguised by a greater variance in scores (which I am sure is the case). For example, if games between two players have a variance of 20 points the komi will have a much smaller influence on the win percentage than if the variance of the scores is 10 points. But this would not necessarily mean that a different komi should be used. In the first case the komi might shift the win percentage from 60% to 50%, in the latter from 70% to 50%, for example.

HermanHiddema: A first statistic for you then, from the gobase database. The results of some 6400 games played in the last 50 years where both players were 9-dan pro, and with at least 5.5 komi:

  • some 3600 end in resignation (B+R or W+R),
  • some 2000 end within a margin about equal to komi (B+6.5 through W+6.5)
  • the remaining 800 end in 7.5 point wins or more (results above +12 are rare, pros will generally resign such games).

On a related note. Of 1409 games played where both players were 9-dan pro and where the komi was 6.5, 709 were won by black, 700 by white. It seems 6.5 is a pretty fair komi.

ekberg: Remember that many of the resigned games are caused by losing a big, complex fight - a fight that might not have occurred if komi were different. So be careful when using the statistics!

cbc: ekberg makes an important point. Strong players count often and adjust their strategies according to whether they are ahead or behind. If we generalize game strategy as playing moves that have the greatest probability of producing a win, being behind will tend to induce more make-or-break plays. So I would guess that the higher the komi, the more games will end in resignation. I agree with RobertJasiek that it makes sense to play with proper komi, even if it doesn't make much difference to our games, so we learn to factor komi into our counting and strategy, anticipating that someday it may make a significant difference.



Komi Go / Discussion last edited by 50.23.115.116 on January 26, 2015 - 08:04
RecentChanges · StartingPoints · About
Edit page ·Search · Related · Page info · Latest diff
[Welcome to Sensei's Library!]
RecentChanges
StartingPoints
About
RandomPage
Search position
Page history
Latest page diff
Partner sites:
Go Teaching Ladder
Goproblems.com
Login / Prefs
Tools
Sensei's Library