Opening problems for AI: Problem 25

For lessons, as well as threads about specific moves, and anything else worth studying.
Post Reply
Bill Spight
Honinbo
Posts: 10905
Joined: Wed Apr 21, 2010 1:24 pm
Has thanked: 3651 times
Been thanked: 3373 times

Opening problems for AI: Problem 25

Post by Bill Spight »

Later in the game, 1620-00-00ap, between Nakamura Doseki (W) and Yasui Santetsu.

White to play.
Click Here To Show Diagram Code
[go]$$Bcm23 White to play. Which way?
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . X . O . . . . . . . X . . . . |
$$ | . . O a . . . . . , . . . . . X X . . |
$$ | . . b 1 . . . . . . . . . . . O O . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . O . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . X X X . . . . . . . . . . . . . . . |
$$ | . O X O X X . . . , . . . . O , . . . |
$$ | . . O O O . X . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]
How should White reply to the press, :b23:? Should White push through and cut, or should White crawl on the 3d line? Nakamura chose one, but the other was 12% better, according to Elf (assuming komi, which the game did not have, OC).
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.
lichigo
Lives with ko
Posts: 199
Joined: Fri Mar 15, 2013 2:47 pm
GD Posts: 0
Has thanked: 4 times
Been thanked: 16 times

Re: Opening problems for AI: Problem 25

Post by lichigo »

I love this series, but I always wonder why not using katago with 0 or 0.5 komi? Why o ly Elfv2?
pwaldron
Lives in gote
Posts: 409
Joined: Wed May 19, 2010 8:40 am
GD Posts: 1072
Has thanked: 29 times
Been thanked: 182 times

Re: Opening problems for AI: Problem 25

Post by pwaldron »

Bill Spight wrote:
Click Here To Show Diagram Code
[go]$$Bcm23 White to play. Which way?
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . X . O . . . . . . . X . . . . |
$$ | . . O a . . . . . , . . . . . X X . . |
$$ | . . b 1 . . . . . . . . . . . O O . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . O . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . X X X . . . . . . . . . . . . . . . |
$$ | . O X O X X . . . , . . . . O , . . . |
$$ | . . O O O . X . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]
My first impression was the cut. After thinking a while...I still want to cut. Crawling will leave the white stone on the left overconcentrated and the pincer stone is going to feel pressure, more so with the black hane looming in the top right.

If white cuts then I would expect both sides to settle themselves and white can aim at getting away first to either the top left or bottom right.
Bill Spight
Honinbo
Posts: 10905
Joined: Wed Apr 21, 2010 1:24 pm
Has thanked: 3651 times
Been thanked: 3373 times

Re: Opening problems for AI: Problem 25

Post by Bill Spight »

lichigo wrote:I love this series, but I always wonder why not using katago with 0 or 0.5 komi? Why o ly Elfv2?
Because the GoGoD Elf commentaries have already done the heavy lifting, with a large number of rollouts per move. I don't have to spend any time waiting for results.

As for 0 or 0.5 komi, that would be valuable, but I am interested here in how to play under modern conditions. :)

Edit: And why bother with pre-komi games? First, because human opening theory is largely based upon best play in the late 19th century, and a lot of ideas go back even further. And, surprisingly, AI often endorse those plays in those games, komi or no komi. Second, because these games and their mistakes are easier to explain. For me, anyway. But that is because I learned 20th century opening theory, which overestimated side extensions from the 4-4. From AI I have learned that they are mistakes, but I really can't explain why. Now, in the 17th century top players may have made mistakes by playing on the side, but those mistakes I can explain. :)

For example, :b23: in this game may be considered as a refutation of White's extension to C-11. It's not, but it's a good try. If White plays the jump attachment at E-16, C-11 becomes a good play. (Not that Santetsu thought that C-11 was a mistake, but he knew that E-16 would have been a good followup for White.) These are things that I learned at Mother Takagawa's knee. ;)
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.
User avatar
Knotwilg
Oza
Posts: 2432
Joined: Fri Jan 14, 2011 6:53 am
Rank: KGS 2d OGS 1d Fox 4d
GD Posts: 0
KGS: Artevelde
OGS: Knotwilg
Online playing schedule: UTC 18:00 - 22:00
Location: Ghent, Belgium
Has thanked: 360 times
Been thanked: 1021 times
Contact:

Re: Opening problems for AI: Problem 25

Post by Knotwilg »

With two stones in support no doubt AI will cut
User avatar
Shaddy
Lives in sente
Posts: 1206
Joined: Sat Apr 24, 2010 2:44 pm
Rank: KGS 5d
GD Posts: 0
KGS: Str1fe, Midorisuke
Has thanked: 51 times
Been thanked: 192 times

Re: Opening problems for AI: Problem 25

Post by Shaddy »

I recall that there's a theme with this side shape that Black can't find a good time to stop pushing White. If he pushes once or twice, hane is severe, three times and the hane is a tiger's mouth, but the fourth push is gote. Also White's push damages the pincer stone, which is unpleasant with the top right shape. I think the computer will crawl.
Uberdude
Judan
Posts: 6727
Joined: Thu Nov 24, 2011 11:35 am
Rank: UK 4 dan
GD Posts: 0
KGS: Uberdude 4d
OGS: Uberdude 7d
Location: Cambridge, UK
Has thanked: 436 times
Been thanked: 3718 times

Re: Opening problems for AI: Problem 25

Post by Uberdude »

Knotwilg wrote:With two stones in support no doubt AI will cut
You mean with that extra stone that will become overconcentrated after pushing and cutting gives black momentum to damage the other one? ;-)
Bill Spight
Honinbo
Posts: 10905
Joined: Wed Apr 21, 2010 1:24 pm
Has thanked: 3651 times
Been thanked: 3373 times

Re: Opening problems for AI: Problem 25

Post by Bill Spight »

I see that there is great interest in a no komi analysis of these ancient games. :) I think that's wonderful, but I think that assuming komi has relevance for our modern understanding of opening moves, even given positions from games of yore.
Click Here To Show Diagram Code
[go]$$Wcm24 Actual game
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . X . O . . . . . . . X . . . . |
$$ | . . O . . . . . . , . . . . . X X . . |
$$ | . . 1 X . . . . . . . . . . . O O . . |
$$ | . . 3 2 . . . . . . . . . . . . . . . |
$$ | . a 5 4 . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . O . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . X X X . . . . . . . . . . . . . . . |
$$ | . O X O X X . . . , . . . . O , . . . |
$$ | . . O O O . X . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]
The crawl, :w24:, loses 12% to the push and cut. Elf has a very slight preference for the keima at a instead of :w26:, but White takes gote either way.
Click Here To Show Diagram Code
[go]$$Wcm24 Push and cut
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . X . O . . . . . . . X . . . . |
$$ | . . O 1 2 . 6 7 . , . . . . . X X . . |
$$ | . . . X 3 4 . . . . . . . . . O O . . |
$$ | . . . a 5 . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . W . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . X X X . . . . . . . . . . . . . . . |
$$ | . O X O X X . . . , . . . . O , . . . |
$$ | . . O O O . X . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]
Of course White chooses the hane for :w30: instead of a play on the left such as a, which would make the :wc: stone too close. Besides, ceding the top side to Black is unappetizing. ;) The :wc: stone will be helpful if Black plays first in that area.
Click Here To Show Diagram Code
[go]$$Wcm30 Two step hane
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . X . O . . . . . . . X . . . . |
$$ | . . O O X . X 1 5 , . . . . . X X . . |
$$ | . . . X O X . 2 3 . . . . . . O O . . |
$$ | . . 7 . O . . 4 . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . 6 . . . . . . . . . . . . . . . |
$$ | . . W . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . X X X . . . . . . . . . . . . . . . |
$$ | . O X O X X . . . , . . . . O , . . . |
$$ | . . O O O . X . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]
Joseki for :w32: is to extend with :w34:, but White continues lightly with a two step hane rather than being pushed towards Black's strength in the top right. After :b33:, :w34: makes a strong, solid connection.

At this point Black has a number of playable moves. If he continues with :b35: in the top left, White secures his group with :w36:.

I think that pwaldron puts it well. The crawl makes the :wc: stone overconcentrated and Black's resultant wall weakens the pincer stone. OTOH, the push and cut works with the :wc: stone, and the contact fight on the top side strengthens the pincer stone. Elf's two step hane is a good tesuji here. :)
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.
dhu163
Lives in gote
Posts: 474
Joined: Tue Jan 05, 2016 6:36 am
Rank: UK 2d Dec15
GD Posts: 0
KGS: mathmo 4d
IGS: mathmo 4d
Has thanked: 62 times
Been thanked: 278 times

Re: Opening problems for AI: Problem 25

Post by dhu163 »

Speculatively:
Current position:
1. Neither seems that nice. It seems worrying to try to fight when B has such a solid shape in the lower left and fairly solid in the upper right. Hence, I prefer b=C15 instinctively.
2. However, is C11 close enough to support the fight. D16 E16 E15 F15 E14 G16 H16?

Opp to play
1. C15=b of course, trying to kill corner.

Past:
1. B ahead as from discussion in problem 24, B is very solid in lower left and W has nothing much yet.

Check:
1. I keep forgetting lower right but it turns out not to matter this time. The idea seems to be that W is alive and B is not.
Post Reply