I think stockfish cheated • page 1/2 • Lichess Feedback • lichess.org

kasishp

#1

I feel like stockfish cheated

https://lichess.org/c4RFJcpQ

Stockfish blundered and called it a draw

I feel like stockfish cheated https://lichess.org/c4RFJcpQ Stockfish blundered and called it a draw

mkubecek

#2

As far as I can see, it's a draw by 50 move rule. Last pawn move seems to be 53. f3, I don't see any pawn move or capture since that.

@kasishp said in #3:

That is only for the first 50 moves

No idea what you want to say here, but this is a clear draw. 50 moves with no pawn move or capture.

@kasishp said in #3: > That is only for the first 50 moves No idea what you want to say here, but this is a clear draw. 50 moves with no pawn move or capture.

tiranor

#5

I would say stockfish is really trolling, blundering the piece at the same time as declaring the draw wtih the 50 moves rule :D

mkubecek

#6

IMHO it's a similar problem as Stockfish playing absurdly weak moves in clearly lost endgames (or drawn that cannot be possibly lost). The engine sees that whatever it plays, the outcome is exactly the same (a draw) so that it doesn't recognize good and bad moves and just plays a (pseudo)random one.

Brian-E

#7

The surprising thing to me is the server analysis calling 103.Bf7 a blunder. The server analysis does normally take the ability to claim a draw into account, because moves in winning positions which would not normally throw away the win but in fact allow threefold repetition of position, are called blunders. This time the ability to claim the draw was not noticed by the server analysis.

mkubecek

#8

Actually, I have seen something similar with threefold repetition few times already: the position right before the last move was evaluated as 0.0 (the best move being one that repeats the position) but the final position (its third occurence) was not. As if the position was evaluated under the assumption that the game is going to continue.

tiranor

#9

I'm not surprised with weird evals from the engine, today it gave me a blunder because in an endgame where the only pieces remaining were the kings and my queen, the eval went from M6 to M7 :D

Mr_Kuryakin

#10

@mkubecek said in #6:

IMHO it's a similar problem as Stockfish playing absurdly weak moves in clearly lost endgames (or drawn that cannot be possibly lost). The engine sees that whatever it plays, the outcome is exactly the same (a draw) so that it doesn't recognize good and bad moves and just plays a (pseudo)random one.

It is not the same. You are describing the dying swan bug. It is not caused by the engine. It is caused by Lichess having stockfish using table bases. The moves are hard-coded.

@mkubecek said in #6: > IMHO it's a similar problem as Stockfish playing absurdly weak moves in clearly lost endgames (or drawn that cannot be possibly lost). The engine sees that whatever it plays, the outcome is exactly the same (a draw) so that it doesn't recognize good and bad moves and just plays a (pseudo)random one. It is not the same. You are describing the dying swan bug. It is not caused by the engine. It is caused by Lichess having stockfish using table bases. The moves are hard-coded.