2

In the following endgame position, from Dautov vs Heinemann, from the book FUNDAMENTAL CHESS ENDING, what happens with perfect play?

[FEN "R7/5pkp/4b1p1/3p4/8/5P2/5KPP/8 b - - 0 1"]

The book says black has some difficult problems to solve (but does not clarify whether they win), while a Stockfish 11 analysis at a depth of 54 plies returns +2.52, which is promising for white. What's going on?

1

This position is typical of those that give computers great difficulty because they must be analysed backwards. White will try (somehow) to gain space and penetrate mostly on the dark squares, while placing his Pawns "to maximum advantage" and looking for an opportunity to sacrifice the exchange. In twenty to thirty moves time. If he is lucky (!) there will be some combinational play that increases his advantage.

The problem is that you cannot arrive at this by analysing to 54 ply. At move 1 you have to conceptualize what the winning idea will be, and play toward it. Otherwise, an evaluation based on general principles (as it must be because there are no immediate tactics) may be optimistic but will not get you there. The situation can be summed up with fair accuracy by saying that the computer does not analyse to depth 54, but instead to depth 1, 54 times and this is not at all the same thing.

A strong human player would deal with this by making moves that "look good" while avoiding for as long as possible decisions that cannot be reversed. They would hope that while doing this, any final play that exists might slowly come into focus. If this position had arisen in a world championship match, I think that decades might be spent trying to arrive at a verdict.

2
  • I tried with Houdini Pro 6.02 x64 mode with highest VARIETY (14 possibilities on first move, 12 possibilities on second move, 11 possibilities on third move and 10 possibilities on fourth move, for both BLACK and White. I went to d=44/72 and I reached a rook plus two pawns vs bishop plus two pawn endgame. I tried from there at d=39,same high VARIETY, and seem a White win. I should get at TalkChess.com to be completely sure, of corse.
    – Stefano
    Sep 25 at 0:29
  • 3
    @Stefano your reply seems to ignore the answers point of this position being difficult for computers. Sep 25 at 13:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.