Life In 19x19
http://lifein19x19.com/

Is Komi standard play?
http://lifein19x19.com/viewtopic.php?f=11&t=10387
Page 2 of 3

Author:  bayu [ Mon Jun 02, 2014 10:38 am ]
Post subject:  Re: Is Komi standard play?

leichtloeslich wrote:
HermanHiddema wrote:
currently OGS always gives the stronger player white

Wow, that's seriously braindead and entirely defeats the purpose of giving white komi.
I hope it's a bug and not a feature.


It is more a feature than a bug. If I remember correctly, it doesn't apply when you challenge somebody. So it affects "only" ladders and tournaments.

leichtloeslich wrote:
illluck wrote:
You can't try to adjust komi after the game - players will play differently with different komi

I'm amazed this needed to be pointed out.

I regularly sacrifice a few points in 9x9 endgames to simplify the game.


I principally agree with you, but if it is just 1 point you might get away with it concerning the amateur games. I don't think you make a 1.5-win deliberately into a 0.5-win because it simplifies the game. And given the data that is there, it's at least something possibly interesting you can look at. I'm no coder, so I'm happy if other people do this kind of things. And I'm even happier when they explain what they do. I simply reserve the right to interpret the results cautiously, if at all. :)

Author:  illluck [ Mon Jun 02, 2014 10:46 am ]
Post subject:  Re: Is Komi standard play?

My understanding is the higher rated player always being white is a bug.

I think the analysis is unfortunatedly too affected by players' ability to count. I think if the winrate is around 50% at 6.5 komi then it's a decent enough indication of it being approximately the right amount at this time. My personal experience is that 6.5 is not too far from being correct.

Author:  Bonobo [ Mon Jun 02, 2014 1:00 pm ]
Post subject:  Re: Is Komi standard play?

I wonder, then, why komi (except on ☯GS and DGS) still sticks with the 0.5 instead of using anywhere between 0.1 and 0.9 (plus whatever may be needed to “guarantee” evenness).

Author:  Dusk Eagle [ Mon Jun 02, 2014 1:37 pm ]
Post subject:  Re: Is Komi standard play?

bayu wrote:
I principally agree with you, but if it is just 1 point you might get away with it concerning the amateur games. I don't think you make a 1.5-win deliberately into a 0.5-win because it simplifies the game.

I have most definitely read 9x9 games to the end following the simple variation, determined I was ahead or behind by 0.5 points, and made my moves based on that. I can even think of a very particular 13x13 tournament game where I read that I would lose by 0.5 points following simple variations in the endgame and had to change up my play to increase my chances of winning, even if in the worst case it cost me more points. It's really not that hard for an amateur dan to do.

Author:  RBerenguel [ Mon Jun 02, 2014 1:39 pm ]
Post subject:  Re: Is Komi standard play?

Dusk Eagle wrote:
bayu wrote:
I principally agree with you, but if it is just 1 point you might get away with it concerning the amateur games. I don't think you make a 1.5-win deliberately into a 0.5-win because it simplifies the game.

I have most definitely read 9x9 games to the end following the simple variation, determined I was ahead or behind by 0.5 points, and made my moves based on that. I can even think of a very particular 13x13 tournament game where I read that I would lose by 0.5 points following simple variations in the endgame and had to change up my play to increase my chances of winning, even if in the worst case it cost me more points. It's really not that hard for an amateur dan to do.


Not even dan, I regularly try to squeeze some extra points in 9x9 to win if I see a close defeat ahead

Author:  RobertJasiek [ Mon Jun 02, 2014 9:29 pm ]
Post subject:  Re: Is Komi standard play?

19x19: Both theoretical studies by me, Bill Spight and others, and professional playing experience suggest 7 +- 0.5 komi. A final proof might be missing ALA the game is unsolved.

13x13: The most meaningful available evidence for 8.5 komi is the greatest likelihood of even-matched 0.5 score games in European 13x13 Championships, as I have observed and experienced 1993 - 2013. The value is greater than for 19x19 because 13x13 also is a lot about killing, and having the first move plays a great role for that.

9x9: Japanese professionals have used 6.5 komi for a longer time than for 19x19.

Author:  RBerenguel [ Mon Jun 02, 2014 10:58 pm ]
Post subject:  Re: Is Komi standard play?

RobertJasiek wrote:
19x19: Both theoretical studies by me, Bill Spight and others, and professional playing experience suggest 7 +- 0.5 komi. A final proof might be missing ALA the game is unsolved.

13x13: The most meaningful available evidence for 8.5 komi is the greatest likelihood of even-matched 0.5 score games in European 13x13 Championships, as I have observed and experienced 1993 - 2013. The value is greater than for 19x19 because 13x13 also is a lot about killing, and having the first move plays a great role for that.

9x9: Japanese professionals have used 6.5 komi for a longer time than for 19x19.


I'd have never thought possible you'd justify something based on what (Japanese) professionals do :P

Author:  RobertJasiek [ Mon Jun 02, 2014 11:24 pm ]
Post subject:  Re: Is Komi standard play?

I prefer best possible justifications, and currently I am not aware of any better justification for 9x9 komi.

Author:  Scurvodsky [ Tue Aug 26, 2014 1:17 pm ]
Post subject:  Re: Is Komi standard play?

Between newer players, i.e. players who don't really understand what they are doing yet, komi seems quite large and appears to give the advantage to white. It would seem to me that only between skilled players would moving first give that big of an advantage.

Author:  emeraldemon [ Tue Aug 26, 2014 3:08 pm ]
Post subject:  Re: Is Komi standard play?

RobertJasiek wrote:
A final proof might be missing ALA the game is unsolved.


This. Imagine a scathing line by black that forces the komi to be 8, or 10! Maybe we just haven't found it yet.

Author:  jug [ Tue Aug 26, 2014 3:50 pm ]
Post subject:  Re: Is Komi standard play?

Dusk Eagle wrote:
If we want to determine what the win rate would be at 4.5 or 5.5 komi, I think the only valid way would be to generate a large sample of games played at 4.5 or 5.5 komi.

Here are some numbers from DGS (sample of ca. 83.000 finished 9x9 games) with counts, percentages of Black-win (BWin) and White-win (WWin) and Jigo (BWin & WWin are shown without Jigo; add half of Jigo-percentage if you want) for a variety of Komi. Time-Outs are not included and each game has at least 10 moves.
Code:
| Komi | CNT   | BWin    | WWin    | Jigo   |
+------+-------+---------+---------+--------+
| -6.5 |   342 | 45.0292 | 54.9708 | 0.0000 |
| -6.0 |   418 | 50.2392 | 48.3254 | 1.4354 |
| -5.5 |   467 | 45.8244 | 54.1756 | 0.0000 |
| -5.0 |   486 | 51.2346 | 47.1193 | 1.6461 |
| -4.5 |   578 | 51.0381 | 48.9619 | 0.0000 |
| -4.0 |   547 | 49.5430 | 47.8976 | 2.5594 |
| -3.5 |   630 | 49.8413 | 50.1587 | 0.0000 |
| -3.0 |   586 | 46.0751 | 51.3652 | 2.5597 |
| -2.5 |   627 | 45.9330 | 54.0670 | 0.0000 |
| -2.0 |   696 | 48.9943 | 48.7069 | 2.2989 |
| -1.5 |   668 | 50.7485 | 49.2515 | 0.0000 |
| -1.0 |   706 | 44.3343 | 52.9745 | 2.6912 |
| -0.5 |   721 | 48.1276 | 51.8724 | 0.0000 |
|  0.0 |  4305 | 47.4564 | 50.8711 | 1.6725 |
|  0.5 | 17091 | 40.2785 | 59.7215 | 0.0000 |
|  1.0 |   868 | 42.0507 | 55.2995 | 2.6498 |
|  1.5 |  1192 | 49.6644 | 50.3356 | 0.0000 |
|  2.0 |   953 | 42.2875 | 54.4596 | 3.2529 |
|  2.5 |  1393 | 48.1694 | 51.8306 | 0.0000 |
|  3.0 |  1016 | 44.7835 | 52.4606 | 2.7559 |
|  3.5 |  2132 | 50.0469 | 49.9531 | 0.0000 |
|  4.0 |  1195 | 45.1046 | 52.7197 | 2.1757 |
|  4.5 |  1735 | 47.0317 | 52.9683 | 0.0000 |
|  5.0 |  1487 | 45.9314 | 51.3786 | 2.6900 |
|  5.5 |  2516 | 48.6089 | 51.3911 | 0.0000 |
|  6.0 |  1589 | 47.8918 | 50.0315 | 2.0768 |
|  6.5 | 28549 | 46.3414 | 53.6586 | 0.0000 |
|  7.0 |  9039 | 46.0892 | 49.2422 | 4.6687 |
|  7.5 |    68 | 48.5294 | 51.4706 | 0.0000 |
+------+-------+---------+---------+--------+
| ALL  | 82732 | 45.4419 | 53.6443 | 0.9138 |

The sample requires probably more restrictions as it seems White have a higher winning-percentage almost regardless of used komi ;-)

Author:  Uberdude [ Tue Aug 26, 2014 3:56 pm ]
Post subject:  Re: Is Komi standard play?

jug, presumably those komis are to handicap games between players of different strengths? What we need is win rates for different komi for games between players of equal strength. (Or different strength but random colour and a much larger sample to see the deviation fron 50%).

Author:  Abyssinica [ Tue Aug 26, 2014 4:19 pm ]
Post subject:  Re: Is Komi standard play?

A lot of people have higher win rates as white.

Author:  Scurvodsky [ Tue Aug 26, 2014 4:44 pm ]
Post subject:  Re: Is Komi standard play?

Abyssinica wrote:
A lot of people have higher win rates as white.


Because the komi helps them or b/c the stronger player usually plays white?

Author:  jug [ Wed Aug 27, 2014 1:50 am ]
Post subject:  Re: Is Komi standard play?

Uberdude wrote:
jug, presumably those komis are to handicap games between players of different strengths? What we need is win rates for different komi for games between players of equal strength. (Or different strength but random colour and a much larger sample to see the deviation fron 50%).

You are right, I didn't restrict the sample to even games. Here's a list from DGS with more restrictions:
* finished games
* more than 10 moves
* size 9x9
* no time-outs
* no handicap stones
* Rating-diff between Black and White <1.5k
* (rated and unrated games, no restriction on strength of players)

Code:
| Komi | CNT   | BWin    | WWin    | Jigo   |
+------+-------+---------+---------+--------+
|  0.0 |   631 | 57.5277 | 41.5214 | 0.9509 |
|  0.5 |  1132 | 59.6290 | 40.3710 | 0.0000 |
|  1.5 |    88 | 51.1364 | 48.8636 | 0.0000 |
|  2.5 |   123 | 60.9756 | 39.0244 | 0.0000 |
|  3.0 |    23 | 52.1739 | 47.8261 | 0.0000 |
|  3.5 |   342 | 56.7251 | 43.2749 | 0.0000 |
|  4.0 |   716 | 47.4860 | 50.0000 | 2.5140 |
|  4.5 |  1035 | 46.2802 | 53.7198 | 0.0000 |
|  5.0 |   903 | 48.3942 | 49.3909 | 2.2148 |
|  5.5 |  1290 | 46.8217 | 53.1783 | 0.0000 |
|  6.0 |   992 | 46.0685 | 51.4113 | 2.5202 |
|  6.5 | 10582 | 46.8626 | 53.1374 | 0.0000 |
|  7.0 |  2472 | 45.1052 | 49.2314 | 5.6634 |
|  7.5 |    24 | 45.8333 | 54.1667 | 0.0000 |
+------+-------+---------+---------+--------+
| ALL  | 20389 | 47.9965 | 50.9736 | 1.0300 |

the query is only for me for reproducing the sample: select Komi, count(*) as CNT, sum(if(Score<0,1,0))/count(*)*100 as BWin, sum(if(Score>0,1,0))/count(*)*100 as WWin, sum(if(Score=0,1,0))/count(*)*100 as Jigo from Games where Status='finished' and Size=9 and Score between -1500 and 1500 and Moves > 10 and Handicap=0 and Black_Start_Rating>-1000 and White_Start_Rating>-1000 and abs(Black_Start_Rating - White_Start_Rating) < 150 group by Komi with rollup having CNT > 20  ;

Author:  Abyssinica [ Wed Aug 27, 2014 5:26 am ]
Post subject:  Re: Is Komi standard play?

Scurvodsky wrote:
Abyssinica wrote:
A lot of people have higher win rates as white.


Because the komi helps them or b/c the stronger player usually plays white?


With any sort of handicap as white on kgs, komi = 0.5

Author:  HermanHiddema [ Wed Aug 27, 2014 6:55 am ]
Post subject:  Re: Is Komi standard play?

@jug: How is it that you have data for all those different komi values? I can't imagine that a nigiri game has random komi, and if the values are as they are to compensate for strength difference, then the data is meaningless.

Author:  emeraldemon [ Wed Aug 27, 2014 8:46 am ]
Post subject:  Re: Is Komi standard play?

HermanHiddema wrote:
@jug: How is it that you have data for all those different komi values? I can't imagine that a nigiri game has random komi, and if the values are as they are to compensate for strength difference, then the data is meaningless.


1) DGS has a "proper komi" option, which adjusts komi based on strength difference. I imagine this is where most of those games come from, and I agree those games don't answer the question.

2) DGS has a "fair komi" option, where players can bid on komi in a few different ways (open auction, secret auction, etc.). These could presumably be good indicators, I don't know how many games are played this way. I tried three myself, bid 8 or 8.5 to play black and won all three :)

3) Of course a player can just put a game in the waiting room with an unusual komi and see if anyone will take. I doubt this happens very often.

Author:  jug [ Wed Aug 27, 2014 1:58 pm ]
Post subject:  Re: Is Komi standard play?

HermanHiddema wrote:
@jug: How is it that you have data for all those different komi values? I can't imagine that a nigiri game has random komi, and if the values are as they are to compensate for strength difference, then the data is meaningless.

I didn't pay attention to what conditions the sample should fulfill to answer some question, but just provided some data because I can ;-) with access to the DGS-data.
But actually ... what was the question again? The above sample-data was only for 9x9 games, ... or are we talking about other board-sizes as well?

The 2nd post with numbers have a rating-diff-restriction, so the players are not more than 1.5kyu apart, but still the majority of games were indeed setup by conventional handicap (fix komi 6.5) or proper handicap taking the rating-diff into account (even though the smaller the board-size the less influence the rating-diff has).

emeraldemon is right about that DGS has this proper-handicap, but DGS also has a way for manual handicap/komi-setup (nigiri, choose black or white color, or start a double-game). However, I just checked and the number of games for the manual setups are probably too small (counts in range 20-100 with less komi-variety) to have some reliable sample-size.

emeraldemon wrote:
2) DGS has a "fair komi" option, where players can bid on komi in a few different ways (open auction, secret auction, etc.). These could presumably be good indicators, I don't know how many games are played this way

There are ca. 300 finished DGS-games that used fair-komi for game-setup, but the majority was for 19x19.

emeraldemon wrote:
3) Of course a player can just put a game in the waiting room with an unusual komi and see if anyone will take. I doubt this happens very often.

If we are only talking about 9x9 games, then there are ca. 10.000 finished DGS-games that were started with a even-game with nigiri and manual komi, but the majority of those were using komi 6.5 or 7 (a lot of those games are the 9x9 ladder-tournament). If we would further restrict the games on the players-strength or max 1k-rating-diff it would be only 25% of those counts.

I'm not sure, if I can be helpful with providing DGS-data ... if it's clear what exactly the restrictions should look like, I may provide some data ... though in this case it doesn't sound like it ;-)

Author:  coconut [ Fri Apr 05, 2019 12:52 pm ]
Post subject:  Re: Is Komi standard play?

I hate komi. I don't think black has an advantage because it goes first. It's a turn based game, both black and white have equal opportunities to make their marks on the board. In the end it's about cleverness and strategy and not who goes first.

I also hate losing just because of Komi points.

I mean, I consider Komi just for tournaments and not a rule. It is a fairly new rule and it's different in every country. In Hikaru no Go, Sai the Go ghost has no idea what Komi is. He's thousands of years old and he knows how to play Go better than anyone alive, but Komi is new to him.

Kind of like how my dad plays backgammon where you have to get the exact number to bear it off every time, you can't start taking off the fours if you get a five and don't have any pieces on the fives or sixes. You have to get a four. He thinks it's a better way to play, and doesn't care that standard rules are different.

I have more fun with my dad's rules anyway.

My dad isn't playing backgammon at tournaments so it doesn't matter. Just for playing regular games for fun, or standard play, one of the rules, I can't consider Komi a real rule because it is a tournament addition.

Page 2 of 3 All times are UTC - 8 hours [ DST ]
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/