Evaluation bug

From baby prodigy to tablebases, everyone agrees that move 61.Kxd8 is better than the other legal move. Position is:

But such a brilliant move got inaccuracy symbol?!

My explanation: Evaluation (from a very, very quick analysis) went down from -25.9 to #-13. This makes it a questionable move according to the settings.

Conclusion: The automatic computer analysis is just a very quick and simple blunder check. Do not take it too seriously.

Funny you mentioned tablebase as it may in fact be part of the problem. As 61. Kxd8 leaves exactly eight pieces on the board (including kings), it means that the position after it can be evaluated using tablebase while both the position before this move and after 61. e5 cannot which may be the reason why SF evaluates it as "only" -25.88 in the server side analysis. When you let local engine run long enough, both the position after 60... Rxd8 and after 61. Kxd8 evaluate as #-11. (And the position after 61. e5 as #-9.)

@mkubecek said in #3:

61. Kxd8 leaves exactly eight pieces on the board (including kings)

8 is too much for current tablebases, it does catch up the position but the number is 7.

@Entuzijasta said in #4:

8 is too much for current tablebases, it does catch up the position but the number is 7.
Oops, I was in fact looking at the position before Rxd8, not after, when I wrote comment #3. (And thought we already have 8-pieces tablebase available for server side analysis.)

Perhaps Stockfish was drunk when analysing that game?!

And again, crazy evaluation for the last move.