How the hell is this endgame WINNING for White????

@petri999 said in #9:
> 8/2Nr4/4Kp2/4pN2/4Pk2/8/8/8 w - - 0 53
> the position I saw on boad and table base gives Kxd7 as winning move. in actual game hitting win Kn is drawn.

Oh I see, you're taking it from before the K+N+N vs K+P endgame was reached.

I maintain that the position after 55.Nxd7 is drawn.

The position at the end (after 56 e3) is indeed a draw according to lichess tables and so is everything starting with 53... Kxe4. However, in the position after 52. Rxd7, the move 53. Kxd7 (which wasn't played) is shown as winning.

This is a stunning position. Thanks for posting!

Here are two games lichess.org/study/mUghItjz/wLm9oFEK I recently witnessed on my own laptop, where I am running an endgame round robin tournament of 16 engines. Note that I am running the tournament without tablebases. In chapter 1, Caissa 1.18 wins against Stockfish 16.1 in a KNN vs kp endgame. The ending is beautiful. Caissa 1.18 won it at 40 moves per minute on an old laptop. In chapter 2, again we have Caissa 1.18 beating Stockfish 16.1, this time a KBN vs k ending. For both starting positions, Stockfish 16.1 won the games with colours reversed, but the endings were different. So it is likely that both the starting positions are somewhat favourable to white.

Hard to believe you can have a 2000 FIDE rating and still be baffled by this sort of thing...

How many pieces can be on the Board for a table base solution. Are 8 pieces not to much?

@e4e5f4exf4nicht said in #16:
> How many pieces can be on the Board for a table base solution. Are 8 pieces not to much?

Seven is the maximum that has been achieved so far (the two kings and five others). Seven-piece tablebases were completed during the 2010s decade.
Eight-piece tablebases are not currently feasible.

There is a small bug on Lichess.
If on move 53 White takes the Rook (stockfish best recommended move with 8 pieces) then white is winning from the tablebase and it could swap from 1.4 pawn advantage to saying how many moves checkmate will be in (34 moves I think)

I'm not sure if lichess software has this information available as the position after 53. Kxd7 (or the three replies to it) only shows as winning for White with "DTZ 3" but it does not show the DTM value. And even if it knew the DTM values it would still require traversing the tree to some extent to be able to show the best line. But I agree that such output looks a bit inconsistent.