Life In 19x19

White wins again. I wonder if anyone has recent stats for white win percentage, maybe komi needs to decrease.

White wins again. I wonder if anyone has recent stats for white win percentage, maybe komi needs to decrease.

Your suspicions seem correct.

Crude figures:

2018: 1657 games, W wins 900 (54.3%)
2017: 4291 games, W wins 2291 (53.3%)
2016: 3557 games, W wins 1803 (50.7%)
2015: 2879 games, W wins 1465 (50.9%)
2014: 2596 games, W wins 1313 (50.6%)
2013: 3815 games, W wins 1923 (50.4%)
2012: 3233 games, W wins 1616 (50.9%)
2011: 2701 games, W wins 1404 (51.2%)
2010: 2735 games, W wins 1364 (49.9%)

A couple of data points from years of smaller komi:

1990: 1431 games, W wins 703 (49.1%)
1985: 986 games, W wins 452 (45.8%)

Conceivably the recent rise in White's improvement is linked with something pros have learnt from AI (and not let on about)?

A crude ctrl+F search using Go4go gives me this (barring any mistakes on my part),

Ke Jie's last 450 games
W+ 263 games (58,4%)

Park Junghwan's last 450 games
W+ 231 games (51,3%)

Fan Tingyu's last 420 games
W+ 224 games (53,3%)

Chinese Mingren's last 1000 games
W+ 517 (51,7%)

Chinese Tianyuan's last 450 games
W+ 247 (54,8%)

Chinese League 1's last 450 games
W+ 249 (55,3%)

Since the games played by each player include many 6,5-komi games, I also looked at Chinese tournaments to have samples of 7,5-komi games only. I don't know how significant these numbers can be be from a statistical point of view but could it be that Ke Jie is significantly better with white (would be nice to have a stat for his 7,5-komi games only but I guess it could only be higher)?

Couldn't we use AI to determine the correct komi for white? I've always wondered about that.

Since the games played by each player include many 6,5-komi games, I also looked at Chinese tournaments to have samples of 7,5-komi games only. I don't know how significant these numbers can be be from a statistical point of view but could it be that Ke Jie is significantly better with white (would be nice to have a stat for his 7,5-komi games only but I guess it could only be higher)?

I have 176 games with Ke Jie as White and 3.75 komi. He won 138 (and 123 without counting).

As Black he played 185, winning 121.

I found his White performance so astonishing (78%) I had to re-check the figures, but I have seen a reference that he likes to play White. His performance as Black was 65%.

If you add in the 6.5 komi games from international events (tougher competition?) his White score drops to 65% for those games only (26 wins out of 40). With Ing komi his White score is 3-3.

Strangely, with 6.5 komi and Black against humans, it seems Ke Jie scored just 14/41 (34%).

Ke Jie’s strength with White has been discussed on these boards before. I believe it predates the AI era, though I wish I had a link.

Elom wrote:Couldn't we use AI to determine the correct komi for white? I've always wondered about that.

You could certainly use AI to estimate the correct komi, and the stronger the AI, the better the estimate would be. They're not yet (and will probably never be) at the level where they could tell us definitively. The "correct" komi appears to be about in the range that humans have agreed upon, which is not surprising.

John Fairbairn wrote:I found his White performance so astonishing (78%) I had to re-check the figures, but I have seen a reference that he likes to play White. His performance as Black was 65%.

Astonishing indeed! It seems to me that not only is he better playing with white than with black but also that he is better playing with white than other professionals are.

dfan wrote:

Elom wrote:Couldn't we use AI to determine the correct komi for white? I've always wondered about that.

You could certainly use AI to estimate the correct komi, and the stronger the AI, the better the estimate would be. They're not yet (and will probably never be) at the level where they could tell us definitively. The "correct" komi appears to be about in the range that humans have agreed upon, which is not surprising.

I'm pretty sure this is untrue, or at least unimportant. It should be much more important that the two opponents are evenly matched. AIs have already shown us that they prefer White with 7.5 komi in Chinese rules, so that's something.

What would be interesting is seeing how AIs play with different komi, e.g. Golaxy since their developers stated that it could play with different values of komi.

The race for No. 1 is getting tight again, as Mi Yuting just beat Park Junghwan by 0.5 point in Chinese A League, while Ke Jie survived Chen Yaoye by 1.5 points.

Today Ke Jie slipped into 3rd on goratings after losing to Gu Zihao and Mi Yuting beating Fan Tingyu in the 15th round of the Chinese A League. Ke Jie, Mi Yuting, and Shin Jinseo are all within 5 points of each other, although this will probably fluctuate since Ke's past ratings are much higher than the latter two.

In Ke Jie's game against Gu Zihao, it seems that he got into a bit of trouble in the center. His center group was split in two, with one half eventually dying. However, after :white:192, my version of Leela Zero (#157, the last 15b network) seems to think there's still a bit of aji and a potential ko on the side after the D13 push for :black:193. Granted, the winning percentage was only in the 30s, but LZ didn't like the exchanges on the bottom. Would a stronger player care to shed some insight onto this? Was Ke Jie simply trying to play a tricky move because he saw the push didn't work, or was it really a misread?

Today Ke Jie further slipped into 4th position on goratings, only because he had a weaker opponent in yesterday's Chinese league. Mi Yuting is only 16 points behind No. 1 Park Junghwan after Park lost to Yi Lingtao. Yi Lingtao is a rising star that we should pay a lot of attention.

This could become very interesting if we have a close four way contest for #1!

Great game between Ke Jie and Mi Yuting today in the Chinese league. Mi showed balls of steel to start a big approach ko which put some of his stones (which could have been safe) in jeopardy to avoid Ke's sente yose, but also put the status of Ke's even bigger corner in question and Mi had enough big threats to win by resign.

What about the alphago joseki in the lower right? forum/viewtopic.php?f=15&t=16104