:: Igo Hatsuyôron 120 (2021)

Variations of Our Solution (IIIb)

Tricking KataGo (Triple-Ko)

VAR (C) (A) (.2) (.2) (.1)

(Referenced by 222fc)


: ( 222bb)
White extends into the corner, aiming at hidden weaknesses.

: (A 222bw)
Black captures White's two stones at the top.

During the investigations on the comparable position with the Guzumi Exchange ( 222go), I found that defending in the corner with the jump to would be much better for him than allowing the ko fight shown below.
However, this is not true here. White will suffer decisively in this ko fight from not having captured the nakade at the left edge yet.

:
White occupies a vital point in the corner.

:
Black peeps from above.

:
White has to connect.


: (A 222bx ; B 222by)
Black also connects, solidifying his position at the upper edge.

During the investigations on the comparable position with the Guzumi Exchange ( 222go), I found that connecting Black's stones on the outside would be much better for him than allowing the ko fight shown below.
However, this is not true here. White will suffer decisively in this ko fight from not having captured the nakade at the left edge yet.

contains a larger threat against White's potential territory at the top, and so prevails over .


:
White gives atari at the left edge.

:
Black must connect immediately, as he is unable to win any ko fight in this corner, due to White's unbeatable ko threats in the nakade and the hanezeki.

:
Now the time is ripe for cutting Black's forces apart.

:
Black pushes into the corner for another time.

Jumping to , instead, will end even worse for him.


: ( A 222bz ; B 222ca ; C 222cb ; D 222cc ; E 222cd)
White's hane in the corner makes it difficult for Black to approach her stones from above.

Please note that here, in the "classical" line of play without the Guzumi Exchange, KataGo (2021i) always favours - independent of the (reverse) komi - which is a remarkable difference to KataGo's evaluation when the Guzumi Exchange has taken place.

We present the subvariations for the other mainly considered alternative options in descendant order of their final score. All of these will end with a White win.

KataGo (2021i) found a decisive improvement for Black in the following subvariation.
White's peep with at the upper edge now also ends with the same optimal score for both sides.

: ( 118 222ce ; B 222dm)
Black pushes at the top, probably influenced by an idea to enter some kind of Joachim's Ko Semeai with White's centre group.

Guided by his experience with this position (at the left) after Black's guzumi at the right, Joachim suggested blocking with in the upper left.


KataGo (2021i) also favours this block, which will result in a much better final score for Black than the sequences below, after KataGo (2021i) found a decisive improvement for Black.


Black's descent to in the corner looks like being too slow, and so played too early. White will choose the Semeai Variation, being able to survice with her corner group in the hanezeki. However, Black would still be better than losing his large right-hand group outright in the Capture Variation, when White chose the correct continuation with the subvariation below.

It seems very noticeable to us that one obviously must not follow KataGo's suggestions without further ado in positions that can only be reached after (multiple) mistakes on both sides. Presumably there is no need for the special neural network to analyse positions of this kind down to the very last detail.

Adjusting the komi helps sometimes but is not a panacea.
Here, for example, spending some additional komi points to Black will result in a priorisation of the "countable" variations (, ) over the desperate measure ().
However, no change can be found in White's preferences. I suspect that the concrete effects will materialise too far in the future, behind KataGo's event horizon.


As a matter of course, Black would be able to capture White's stones in the upper left corner by playing the atari of .
But he must not do so, because he will have to take White's stones really off the board, before he could occupy all of White's liberties in the upper left. Losing a decisive tempo in the Main Semeai is out of the question, as the Punishment Semeai is already lurking around the corner.


: (121 222do ;  C 222dr)
White attacks Black's stones in the corner.

She might consider blocking at , instead, saving her stones at the top. She will lose only a few points, compared to the sequence below.

However, Black's previous move lost one tempo in the Main Semeai, so White's best option is to occupy a liberty of Black's large group at the right, instead.


:
White connects, so capturing Black's stones below.

:
Black gives atari at White's three stones below.

:
It is best for White to connect.

She might consider capturing the hanezeki's tail with , instead, also capturing Black's cut stones in the upper left corner. She will still be better than in the solution, but her win will not be as convincing as here (similar to the subvariation for ). Her gain in the upper left corner will roughly compensate for the loss at the right, caused by the dissolving of the hanezeki.

:
Black defends his territory in the left centre in sente.

:
Black captures White's stones at the top.

:
It follows an amateurish idea of the subsequent endgame.

:
Black must take White's four stones a liberty, because his cut-off large group at the right has only one eye.

White will win the game by 19 points.

Copyright © 2021 Thomas Redecker.

Design by Jan van Rongen, modified by Thomas Redecker.

L e g a l   n o t i c e