# Segoe Tesuji Problem 2

Difficulty: Advanced   Keywords: Tesuji, Problem, Books & Publications

As an example of the tesuji problems in this book not always doing what you would expect (this is an A-difficulty problem), consider the following problem from volume 3 (2 in the old one), from the section on getas:

Segoe Geta Problem #3: Black to play

It appears that black can capture the four white stones in a net, as follows or with a similar loose move.

Black Captures

That is not the given answer, however. This is:

Even though black can capture the four white stones with a looser net, he opts for a tigher net knowing that he can squeeze for an even *better* result if white tries to get out. Remember, these are half-board problems. To get a good result, it is not always a simple "capture the stones" problem.

Interesting, there is much to learn from this kind of problems, I'm sure..But in this case, it seems like a trick play. If w reads it out, he will not try to escape, but will play on the right side instead. Then b still needs another move to capture?

Segoe Geta Problem #3: Black to play

hnishy: The old edition on my bookshelf has two marked White stones. A little different situation.

xela: I have the 2003 edition, and it also has those two white stones.

kmr - here is LZ analysis (chinese rules, komi 7,5 as always with LZ). Surprisingly, LZ strongly disagrees with Segoe.

xela: Since you asked (in the edit summary), I checked with KataGo. Same. By amateur standards, the book solution isn't a blunder (maybe a 3 point loss), but KataGo thinks it's clearly worse. And I tried some different full board positions. It doesn't matter if the two "extra" white stones are there or not. I'm not actually that surprised. I've seen the same thing reviewing older pro games with KataGo. We already know that it tends to value corner territory a bit more highly, and outside influence a bit less, than humans (although there are certainly some positions where it will sacrifice territory for influence). The difference between the two variations here is something like 35 points of territory.

Perhaps it's a case where making this sort of grand gesture (sacrifice one big corner, dominate the rest of the board) will intimidate your human opponents and win games, but it's not such a good idea between two evenly matched superhuman entities?

Segoe Geta Problem #3: Black to play

kmr: If we assume some stones in upper side of board to make it like even game, then in such a position surprisingly LZ thinks that correct move is amateurish "a" (WR for white 25,5%) and "b" is huge mistake (WR for white 63,7%).After black a, LZ thinks that white should play some forcing moves starting with "c", after which white should tenuki and play approach in upper left, resulting in WR=26,1%

Segoe Geta Problem #3: Squeeze variation

After whole sequence is played, LZ evaluate black "5" as mistake, resulting in WR=63,9%, and preferring it at "a" with result of WR=54,4%.If black playes, Segoes "5" LZ responds with "a" which is forcing move that requires some defense (apparently LZ thiunks that black must not allow white to have option to pull his cutting stone, otherwise black is in even bigger troubles).

Black Captures

hnishy: An important factor is how safe the marked stones are. After the loose net in the diagram, Black can't prevent the connect underneath without allowing escape of the netted stones (White may play at as a probe). Segoe's tight (quasi-)net gives White fewer options and is easier to read for humans.

kmr - Yes, white has this nice option, but the loss of having to play is quite big, and this 2 marked stone are not in big danger. Even if black will both play hasami at "a" and deny white conection underneath, its 2 moves - white can just saricifice those stones or treat them lightly. For sure Segoes var is easier for human, and also create very strong pressure on opponent. Without AI, i am pretty sure most of pros (and i guess overwhelming most) would swiftly agree with Segoe.

Segoe Tesuji Problem 2 last edited by kmr on February 19, 2023 - 11:42