Komi pie, lengthening the title? Go forum
10 replies. Last post: 2016-12-03Reply to this topic Return to forum
David J Bush at 2015-10-10
What do you think of this adaptation of the pie rule (used in Hex and Twixt) for Go? Instead of placing a stone to start the game, one player chooses a komi value. Then the other player chooses which side to play. This makes more work for the players, but results in a game which is “future proof” with regards to the appropriate komi for a given size grid.
mmKALLL at 2015-10-10
In my opinion, since komi changes are very infrequent by the related associations, it wouldn’t be necessary to add a pie rule here. In games like Gomoku, Hex and Polyomino, the pie rule creates a lot of interesting opening concepts and positional analysis problems, but I doubt that having one for komi increases the depth of the game. After all, there’s a theoretical exact value for komi, and it’s trivial to pick a preference. Having a pie rule for the initial position on the other hand would be interesting as a variant, but personally I don’t think that’s a high priority either as Random100 exists.
David J Bush at 2015-10-10
"there’s a theoretical exact value for komi, and it’s trivial to pick a preference." I’m not sure what you’re saying. Who picks this preference? Do you know what this exact value is? How does one conclude, from the existence of an exact value, that komi pie would not increase the depth of the game?
mmKALLL at 2015-10-11
I meant preference for a player’s point of view - (nearly) regardless of opponent, I would always play Black if the komi offered was below 6.5 points and White when it would be above 7. For the values inbetween, I’d be fine with either. Picking a preference for color regarding the offered value of komi is not very difficult or interesting.
As for the theoretical value, there has been quite a lot of research regarding the value of each move and the theoretical advantage of the first player to move in Go. My impression is that most professional players feel that 6.5 might be too low for komi (on 19x19), but 7.5 is too high. The statistics of modern professional play seem to hint in that direction as well. Of course, there are slight differences between countries/rulesets as well – under Area Scoring, the player to fill the last dame gets an extra point, so on average Black’s advantage of playing first is increased by half a point, and the method used for calculating scores in China restricts the possible values of komi there.
Sighris at 2015-11-23
Very interesting conversation and ideas. Based on what mmKALLL wrote (on 2015-10-1) "My impression is that most professional players feel that 6.5 might be too low for komi (on 19x19), but 7.5 is too high. The statistics of modern professional play seem to hint in that direction as well..." /end-quote i wonder if people would be open to the idea of playing with a komi of exactly 7 which would allow for a tie game (equal score). I know that in professional games and in tournaments where a reward is given to the person who wins the tournament this causes some extra work (and problems too?) but I think it would be awesome to have ties! And (IMO) a possibility of a “tie” fits in well with the “philosophy of the WeiQi / Go game” (as compared to a Chess game which is more of a “winner takes all” type of game). Does anybody agree with me? Any other thoughts on Komi? Does anybody know if there has been any decent analysis on the komi for 13x13 &/or 9x9 Go games?
hyperpape at 2015-11-23
Komi pie does not meaningfully increase the depth of the game, it doesn’t future proof the game in any meaningful way (if consensus about appropriate komi changes, you simply update the accepted komi), and a few go players appreciate the idea, a larger number find it silly: http://lifein19x19.com/forum/viewtopic.php?f=10&t=8243&hilit=komi+pie.
It’s a natural idea to suggest, but it’s not an idea worth implementing.
mmKALLL at 2015-11-25
I don’t have anything against ties in go; rather, as long as the game is fair, everything is fine.
Besides, in my opinion having a tie after a tight battle is a satisfying result, as it means that both players played well but still ended up as equal. Of course, high-level games tend to end with a half-point difference more often than low-level ones, so I can’t really comment on how often ties will actually occur.
Corrin Lakeland at 2016-10-31
Personally I’m a big fan of having a komi of 7 where ties result in draws. I am somewhat biased, that’s my country’s national rules (NZ). Our experience has been that black wins slightly more games than black, but not enough for us to put the komi above 7. Because of area counting it is very hard for a game to be won by an odd number of points (it requires a shared dame that neither player can play), so we would essentially have to make komi 8 (eliminating draws) or 9, and the majority opinion is that 9 points is perhaps too much.
Approximately 5% of my 9x9 games result in a tie. On 19x19 I’d guess the percentage is similar but slightly lower. In a tournament of 40 games it is rare to have no draws.
Note that ties have a slight impact on play. I had a situation this morning where I was offered a ko during yose. I counted that if I declined the ko then the game would be a draw. I then attempted to count ko threats and concluded that I had more ko threats than my opponent and so I started the ko. I can’t count ko threats exactly so was seriously considering just taking the draw (0.5 tournament points) rather than going for glory. Ties also help out quite a lot with tournaments as they make it less common for SOS to be equal between the two players on the highest score.
Sighris at 2016-11-29
Corrin Lakeland (2016-10-31), very interesting, thanks for sharing. I think you meant: "Our experience has been that black wins slightly more games than WHITE, but not enough for us to put the komi above 7.
Let me see if I follow the rest correctly, since in "area counting it is very hard for a game to be won by an odd number of points" usually Black and White will have a score difference of an even amount, so a komi of a whole number odd amount (such as 7) would rarely result in a tie; however a komi of a whole number even amount (such as 8) could result in a tie; and I would argue should result in a tie in a significant number of games between players of equal (relatively high) skill – that is if a komi of 8 was fair compensation, if however fair compensation for moving 2nd (playing White) is 7.5 then I would expect in a large data-pool of games played with a 7.5 komi between equal skilled pro level players there would be an equal number of White and Black wins, with a significant number of games won by only 1/2 point (and of course no ties since it is impossible to claim 1/2 a point of territory)... Does all of that make sense?
Hmmmm.... I wonder if Google’s AlphaGo (now claimed to be two stones stronger/improved) can help us determine the best (most fair) komi value? It seems to me it could offer up some strong evidence (if allowed to play itself in enough games with various komi settings)!!!
Luis BolaNos at 2016-12-03
"usually Black and White will have a score difference of an even amount, so a komi of a whole number odd amount (such as 7) would rarely result in a tie; however a komi of a whole number even amount (such as 8) could result in a tie"
You are missing the fact that the board size is odd. So an odd komi results in some ties and an even komi almost never results in a tie.