Park Junghwan is The Best

[emeraldemon]

Just to update this thread, Park Junghwan won the 2nd World Go Championship. He won against both Ke Jie and Iyama Yuta. His eight-game win streak was broken by Fan Tingyu in Chinese A League: http://www.go4go.net/go/games/sgfview/73102

[Uberdude]

Although Park lost this one by half a point, I thought this game from today in the Chinese league was excellent. His opponent was Mi Yuting who is on fine form and looks set to overtake Ke Jie to get the #2 spot on goratings.



P.S. can anyone explain f11 and g11? Is there are ladder somewhere with k13 stones?

[macelee]

P.S. can anyone explain f11 and g11? Is there are ladder somewhere with k13 stones?

Cannot explain this exchange. But I can confirm this actually happened - I just checked with somebody who was following the game at the site.

[Uberdude]

Thanks mace, and for getting all those games into go4go so quickly! My current ideas:
- white prepares an attack on black left group (e.g. d10/11) and black's attach is some fancy answer that means he can fight back with e11 or something whilst also having effect on k13 fight.
- preparing j12 or k12 to hem that black group in and try to force it to connect passively at l17. Why doesn't black answer at k12 or thereabouts, worries about left group?
- prepare cutting at l17, though that seems pretty overplay given white's weakness, would expect white to prefer to make those stones stronger and black grovel connect.
- could there be a ladder from n7 area?

Will see what bots say later.

[emeraldemon]

I haven't been following all the new corner strategies, is that sequence in the upper-left new? looks so strange.

$$Bcm9
$$ +---------------------------------------+
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . 3 . . X . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . . . X . . . |
$$ | . . . . 1 . . . . . . . . . . . . . . |
$$ | . . . O 2 . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . O . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . . . X . . . |
$$ | . . . . . X . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ +---------------------------------------+

Click Here To Show Diagram Code

[go]$$Bcm9
$$ +---------------------------------------+
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . 3 . . X . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . . . X . . . |
$$ | . . . . 1 . . . . . . . . . . . . . . |
$$ | . . . O 2 . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . O . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . . . X . . . |
$$ | . . . . . X . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ +---------------------------------------+[/go]

also looks strange to me. The game sprawls out into a crazy fight after black invades. I looked for this sequence on alphago teach and found the same with a different

$$Bcm9
$$ +---------------------------------------+
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . 3 . . . . . . . . . . . . . . . |
$$ | . . a . . X . . . . . . . . . . 4 . . |
$$ | . . . O . . . . . , . . . . . X . . . |
$$ | . . . 2 1 . . . . . . . . . . . . . . |
$$ | . . . O . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . O . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . . . X . . . |
$$ | . . . . . X . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ +---------------------------------------+

Click Here To Show Diagram Code

[go]$$Bcm9
$$ +---------------------------------------+
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . 3 . . . . . . . . . . . . . . . |
$$ | . . a . . X . . . . . . . . . . 4 . . |
$$ | . . . O . . . . . , . . . . . X . . . |
$$ | . . . 2 1 . . . . . . . . . . . . . . |
$$ | . . . O . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . O . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . . . X . . . |
$$ | . . . . . X . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ +---------------------------------------+[/go]

I like just connecting for white, it matches my intuition. I would probably have played at a instead of tennuki.

[Uberdude]

The peep in response to the one point jump is the standard joseki for bots these days if not tenuki (it wasn't unheard of from humans before but was pretty rare / special case move). AlphaGo Master played it a few times and Leela Zero and Elf like it too. Connect is the simplest response (and often good/best) and then if black slides white's 3-3 is unsatisfactory because, had black slid before peep, then the connection in response to peep would be slack (as 3-3 already defends the push and cut). So after slide probably black will 3-3 later and then you wonder how safe the white stick is and if white tenukis again then probably black will pincer and attack it (there are downsides to the peep too, for example after black's 3-3 the b15 slide is less good as white can attach and black's stand is only a single peep not double peep as white already connected from outside peep). The white push up is trying not to get pushed around and makes it less easy for black to view the peep as a light kikashi (if he just sacrifices it then white grew the left side), and if slide then white can not feel so sad to 3-3. But it leaves cutting weaknesses, which black exploits with the 3-3 invasion and a fighting joseki erupts. Ke Jie also played the push up recently so it seems the top pros in China are exploring this move.

[Elom]

The peep in response to the one point jump is the standard joseki for bots these days if not tenuki (it wasn't unheard of from humans before but was pretty rare / special case move). AlphaGo Master played it a few times and Leela Zero and Elf like it too. Connect is the simplest response (and often good/best) and then if black slides white's 3-3 is unsatisfactory because, had black slid before peep, then the connection in response to peep would be slack (as 3-3 already defends the push and cut). So after slide probably black will 3-3 later and then you wonder how safe the white stick is and if white tenukis again then probably black will pincer and attack it (there are downsides to the peep too, for example after black's 3-3 the b15 slide is less good as white can attach and black's stand is only a single peep not double peep as white already connected from outside peep). The white push up is trying not to get pushed around and makes it less easy for black to view the peep as a light kikashi (if he just sacrifices it then white grew the left side), and if slide then white can not feel so sad to 3-3. But it leaves cutting weaknesses, which black exploits with the 3-3 invasion and a fighting joseki erupts. Ke Jie also played the push up recently so it seems the top pros in China are exploring this move.

Interesting. It seems to be another case in which AI is not afraid to settle positions early in the game whereas human pros where timid to do so. I remember one professional saying that it was, in a way, a masterful way to play.

[hyperpape]

It seems to me that it avoids settling, because white does not want to play the safe 3-3.

[Uberdude]

Is "settle" being used with multiple meanings? I would say this peep "cashes the aji" (a play on "aji keshi") or "settles the shape" (somewhat) in that black has many ways to play against the shape of the one point jump (e.g. attaching to either stone, peeping either side, 3-3 invade, tenuki etc) and strong humans tend to be averse to committing to one option which may be wrong (playing aji keshi) so if it's not clear what the best way to play is they have a tendency to play nothing. But if black peeps and white connects is that a soft answer, a good exchange for black, a kikashi? Maybe later white would play a stronger resistance (or tenuki if you play it say as a ladder maker when there is some big ladder going on). Or was it aji keshi, ruining other choices? It's a really hard decision, but in general bots do seem to lean more towards the former whilst humans towards the later (not that humans are homogenous, some might like to make early exchanges too). hyperpape seems to mean "settle the group" as in make the groups safe by securing eyespace (such as with the 3-3 point), so with these short josekis resulting in not yet settled shapes that didn't happen.

[xiayun]

Although Park lost this one by half a point, I thought this game from today in the Chinese league was excellent. His opponent was Mi Yuting who is on fine form and looks set to overtake Ke Jie to get the #2 spot on goratings.

Look like Mi has to wait a bit longer as he lost to Kim Jiseok, while KJ beat Tan Xiao in the latest round. Still, having 3 players with 3,600+ Elo rating (and 2 more on the cusp of joining them) is quite exciting. Really hoping they get to meet each other (Park/Ke/Mi/Lian/Shin) in one of the best-of-5 finals at some point.

[hyperpape]

Thanks Uberdude, I wasn't aware of the two senses that "settling" can have here, and was thinking of settled groups.

[xiayun]

Suddenly Park Junghwan has lost three in a row, and his lead over Mi Yuting at GoRatings is down to 7 points. In fact, there are only 14 points separating the top 4 players. Should be exciting to see who will end the year at No. 1. Too bad Mi isn't getting the same opportunities as the other 3 in international tournaments.

[xiayun]

With Park’s latest loss (to Xie Erhao again) in Chinese A League and Mi Yuting’s win, we may just have a new No. 1.

[Uberdude]

Yeah, Mi Yuting is doing really well recently and his games are great to replay, such a shame he wasn't selected for several big international tournaments recently. Looking at the last 6 months he has 23 wins and 5 losses (86%) and this is not beating up lower grade pros but stuffed full of top players like Ke Jie, Park Junghwan, Shi Yue, Gu Zihao, Chen Yaoye etc. For comparison the others in goratings top 4: Ke Jie has 27 wins, 15 losses (64%), Park Junghwan has 27 wins, 12 losses (68%) and busy Shin Jinseo has 47 wins (including 18 win streak, recently ended by Park), 12 losses (80%). And for comparison Shi Yue who has been maintaining his rating and position around #10 with some good, some bad results has 25 wins, 17 losses (59%). And poor old Lee Sedol who has dropped down to goratings #36 since being #5 back in March 2016 for the AlphaGo match has 16 wins, 26 losses (38%).

[macelee]

Actually, Park Junghwan dropped to No 3 in GoRatings. He also fell below Shin Jinseo by a tiny margin of 0.23 point.