dhu163 wrote:215 losing move. Deepzen's endgame ... this loses 1 point compared to S1. As it loses sente. After top is played, after S1, W must play A4, and B gets Q1 for B4 in sente, and then B wins by playing the only move L19 so that B still gets the last endgame. Most amateur players would miss this mistake.
215: R-01 is a losing sente, losing 3/4 pt. from the original position. So it is the obvious suspect.
Also, Black 231, S-01 instead of A-03, sacrifices 1/6 pt. by allowing White A-04, in order to get B-05, which gains 2 pts. (White has sacrificed 3/4 + 1/6 = 11/12 pt. in these maneuvers!
) By contrast, the alternative sequence given above lets White get B-05 and gain 2 pts. The swing is 4 pts. in the bottom left. However, the swing in the bottom right is also 4 pts., for a wash in the two regions. The difference is who plays first (and last) at temperature 1. In the game it was White, in the alternative sequence it is Black. There is your 1 pt. difference.
However, for Black to play S-01 at move 215 means that White 216 plays first in the top half of the board. And that certainly gains more than 1 pt. It would be quite unusual for a losing sente that sacrifices 3/4 pt. to be correct at this stage of the game. So, yes, Black 215 is probably the losing move. But it still could offer the best chances of winning as a sacrifice to let Black play first in the top right. If so, the losing play comes earlier in the game.
Peng Quan's analysis of 215 seems correct to me, but if so, this is a serious issue with Zen's endgame.
I continue to maintain that top humans have an advantage against bots in the endgame, unless the bots can build a large enough game tree. By the mid-endgame the board has divided into independent regions (except for ko fights) and taking that into account, which is what humans do, is more efficient than whole board reading per se.
.
