I think this should be the last human tournament that has AI participate in it regardless of the outcome.
But I'm not the one who pay the money to the winner, to run the event, so... I'll just watch it quietly here

My initial feeling was also against letting bots use their latest version in tournaments that span a long time, but I suppose the humans are allowed to improve too (but not replace themselves with their twin/child/friend, so to what extent is a new version of DeepZen the same entity?).

Anyway, the opening of Chen Zijian vs. An Joyeong was rather puzzling to me. Following black's knight move at 19 An cut at 20, which is a standard thing to do in the local shape.

Click Here To Show Diagram Code

[go]$$Bcm19
$$ +---------------------------------------+
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . 1 . . . . . . . . . . |
$$ | . . . . . . . . . . O . . . . . . . . |
$$ | . . . . . . . . . X O . . . . X . . . |
$$ | . . X X . . . . . X . . . . . . 2 . . |
$$ | . . O , . . . . . , . O . . O , X . . |
$$ | . . . O . O . . . X . . . . O X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ +---------------------------------------+[/go]

Later An pushed at 28 below, so Chen obviously played 29 in sente and white's shape suffered. Isn't it better for white to play at 29 himself first: if black blocks we usually think that's a good exchange as it means black can't increase his eyespace in sente like he did in the game, and means white can live by connecting there instead of crawling for ever, which also means black's corner isn't alive (locally it's a ko, but p2 hane and connect is sente for cutting point). Did An not like black ignoring and blocking at 28 if he played at 29? Taking the corner is ok isn't it? Maybe An meant to play there first but did the 1-2-3 mistake of imagining he had played the 1-2 exchange when he hadn't and playing 3 straight away? Or if white had previously cut at q7 (but here black has ladder) then crawling instead makes sense because if black plays in the corner then you can play one or more ataris to come up off the second line in sente before connecting.

Click Here To Show Diagram Code

[go]$$Wcm28 An Joyeong (white) vs Chen Zijian
$$ +---------------------------------------+
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . X . . . . . . . X 1 . |
$$ | . . . . . . . . . . O . . . . . X O . |
$$ | . . . . . . . . . X O . . . . X O 3 . |
$$ | . . X X . . . . . X . . . . . X O . . |
$$ | . . O , . . . . . , . O . . O O X 2 . |
$$ | . . . O . O . . . X . . . . O X X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ +---------------------------------------+[/go]

So what happened is An had to crawl, and then played the descent. Apparently this was sente as black answered, I didn't yet read how to win the semeai if black turns on the outside instead. If white can get this point to stop black from living here (like what normally happens) I can see that is a plus for white.

Click Here To Show Diagram Code

[go]$$Wcm36
$$ +---------------------------------------+
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . , . . . . . , . . . . . , X O . |
$$ | . . . . . . . . . . . . . . . . X O . |
$$ | . . . . . . . . X . . . . . . . X O . |
$$ | . . . . . . . . . . O . . . . . X O . |
$$ | . . . . . . . . . X O . . . . X O O . |
$$ | . . X X . . . . . X . . . . . X O . . |
$$ | . . O , . . . . . , . O . . O O X X . |
$$ | . . . O . O . . . X . . . . O X X . . |
$$ | . . . . . . . . . . . . . . 1 . . 2 . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ +---------------------------------------+[/go]

And then he jumped and crawled on the 2nd line some more, and eventually black came back to live. Isn't this an utter failure for white, with the kind of 2nd line crawl we scold 20 kyus for playing? I know AlphaGo famously crawled on the 2nd line against Park Junghwan, but this seems far worse than that, the black wall is hardly weak (though there is gap at a and cut at b to exploit). An then jumped out at 52, which is the move he probably wanted to do before starting all this sequence, and does somewhat reduce the influence of the huge wall he just gave to black. Thinking about the evaluation from Park's game, white has 7 real points (though the group died from a ko threat later), and black doesn't have many real points yet (top right corner fairly open with that peep stone and open on 1st line), but he only needs to get a territory of width 1 along the Q column next to his wall to get as many as white did.

Click Here To Show Diagram Code

[go]$$Wcm50
$$ +---------------------------------------+
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . O . . |
$$ | . . . O . . . . . , . . . . . X X X . |
$$ | . . . . . . . . . . . . . . . . X O 1 |
$$ | . . . . . . . . . . . . . . . . X O . |
$$ | . . . . . . . . . . . . . . . . a O . |
$$ | . . . . . . . . . . . . . . . . X O . |
$$ | . . . . . . . . . . . . . . . . X X O |
$$ | . . . , . . . . . , . . . . . , X O . |
$$ | . . . . . . . . . . 3 . . . . . X O . |
$$ | . . . . . . . . X . . . . . . . X O . |
$$ | . . . . . . . . . . O . . . . b X O . |
$$ | . . . . . . . . . X O . . . . X O O . |
$$ | . . X X . . . . . X . . . . . X O . . |
$$ | . . O , . . . . . , . O . . O O X X . |
$$ | . . . O . O . . . X . . . . O X X . . |
$$ | . . . . . . . . . . . . . . O 2 . X . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ +---------------------------------------+[/go]

Maybe white was afraid of black's turn at white ? It seems not so easy for white to capture the black corner cleanly.

Click Here To Show Diagram Code

[go]$$Wcm28 This results in a ko
$$ +---------------------------------------+
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . X . . . . . . . X 2 . |
$$ | . . . . . . . . . . O . . . . . X O . |
$$ | . . . . . . . . . X O . . . . X O . . |
$$ | . . X X . . . . . X . . . . . X O . . |
$$ | . . O , . . . . . , . O . . O O X 1 . |
$$ | . . . O . O . . . X . . . . O X X 3 8 |
$$ | . . . . . . . . . . . . . . . . 6 4 5 |
$$ | . . . . . . . . . . . . . . . . . 7 . |
$$ +---------------------------------------+[/go]

Apparently Ryan Li and Stephanie Yin just got married.

I heard that too on the AGA stream, but as they only announced their engagement on facebook a week ago I wonder if that got Chinese-whispered into already married, or maybe they did have a very short engagment.

Some results:
- Ryan Li beat Chen Yaoye !!!!!!!!!!
Huge congratulations Probably the biggest ever upset by a Western player, over Hans beating Yoda, Dinerstein beating O Rissei or Aguilar's wins.
- Zen lost! so our musings about changing version are moot
- Ke Jie won fairly quickly with a big kill
- Japan out.
- Park Junghwan, Lee Sedol, Tan Xiao through, Choi's winning streak ends after huge fighting vs Lee.

1. Huang Xin 4p (CN) vs Zhang Xuan 8p (CN)
2. Choi Cheolhan 9p (KR) vs Lee Sedol 9p (KR)
3. Chen Yaoye 9p (CN) vs Ryan Li 1p (CA)
4. Tong Mengcheng 6p (CN) vs Park Yeonghun 9p (KR)
5. Chen Zijian 4p (CN) vs Lu Minquan 4p (CN)
6. Han Taehee 6p (KR) vs Fan Yunruo 6p (CN)
7. Wang Haoyang 6p (CN) vs DeepZenGo (JP, AI)
8. Tang Weixing 9p (CN) vs Tan Xiao 8p (CN)
9. Gao Yu 3p (CN) vs He Yang 1p (CN)
10. Huang Jingyuan 3p (CN) vs Ke Jie 9p (CN)
11. Xie Ke 3p (CN) vs Li Zhe 6p (CN)
12. Guo Yuzheng 5p (CN) vs Li Xuanhao 6p (CN)
13. Kono Rin 9p (JP) vs Rong Yi 5p (CN)
14. Wang Tao 6p (CN) vs Takao Shinji 9p (JP)
15. Liao Yuanhe 5p (CN) vs Na Hyun 8p (KR) N.B. Na 8p now
16. Park Junghwan 9p (KR) vs Zhou Ruiyang 9p (CN)

Again rather strange loss by DeepZen. When a bot lose 0.5 point it feels like it just misread the score.

DeepZen lost. Ryan Li win.
This is basically 'Anything is possible' in Fan Hui voice.
Ke Jie game is very strong, ending very very early, his 10th consecutive wins.

Stayed up all night to watch, worth it!

Amazing win for Ryan Li!

For Zen, is it possible that it calculates the score with Japanese rules whereas this tournament uses Chinese rule? It's very strange for a bot to lose by 0.5!

In the World Go Championship in Japan they said it messed up in the endgame as its value network was trained on Chinese rules/komi but the tournament was Japanese, so it would be perversely stupid if the version for this tournament (using Chinese rules) was designed for Japanese!

Maybe it is their latest version, designed to train the National Japanese Team, so they do it with Japanese rules? Yeah, but I agree it would be pretty stupid to use it in this tournament. Let's see if we have any explanation from Kato Hideki.

Was Chen's cut of 272 a mistake? A timesuji which Ryan ignored to make the seki in the bottom left? Or trying to generate more ko threats? Really impressive that Ryan dealt with that and the nerves/excitement of a possible win vs a top pro in byo-yomi. Maybe all those KGS blitzes helped after all!

AGA stream is up at https://www.youtube.com/watch?v=XHT-rbD1Ieo.

Japan participants are all eliminated but not AGA pro. Interesting!
I'm rooting for Ke Jie and Lee Sedol.

I think DZG's performance today was ... awful.
Its opening was successful might be a masterpiece. But multiple bugs were triggered after W74 and ~W164. The game seemed over after that. White made some minor mistakes in the endgame but the result didn't change. However DZG's endgame was not so bad, esp compared with the games against Park and Mi.

round of 16

Huang Xin 4p (CN) vs Lee Sedol 9p (KR)
Rong Yi 5p (CN) vs Liao Yuanhe 5p (CN)
Xie Ke 3p (CN) vs Tan Xiao 8p (CN)
He Yang 1p (CN) vs Chen Zijian 4p (CN)
Fan Yunruo 6p (CN) vs Wang Tao 6p (CN)
Wang Haoyang 6p (CN) vs Park Yeonghun 9p (KR)

Ke Jie 9p (CN) vs Park Junghwan 9p (KR)
Li Xuanhao 6p (CN) vs Ryan Li 1p (CA)

Wow! Congratulations Ryan! It was really an exciting game.

a western pebble
slung across the mighty seas
brings down a champion

_________________
My days of not taking you seriously are certainly coming to a middle

That was definitely one of the most amazing results ever. Congrats Ryan!

I fell asleep and woke up to this! What a round! Congratulations to Ryan! And Stephanie! It appears they got the certificate the day before this win in China, according to Sina interview.

What a great win for Ryan and the AGA. I didn't expect this, but I'm thrilled with the result!