Sorry to post twice in a row. I just want to precise that the problem happens, at the 15th white's turn, at the very end of the variation of the computer analysis.
Indeed black can mate white in 2 moves : queen e7, white captures with rook, black recaptures with rook -> checkmate.
Consequently best position for the computer is being mated in 2 ? really ?
Indeed black can mate white in 2 moves : queen e7, white captures with rook, black recaptures with rook -> checkmate.
Consequently best position for the computer is being mated in 2 ? really ?
Nice find! (You wrote Qe7 but it's clear you meant Qe1+).
Perhaps the post-game analysis is not "full-throttle" stockfish (to conserve resources)? Or is this a bug?
Perhaps the post-game analysis is not "full-throttle" stockfish (to conserve resources)? Or is this a bug?
Thank you ! You are right, I meant e1 check. Sorry ^^'.
I used stockfish 6 in Tarrash and set up the same position and the engine gave me this line 1.Qa7 Qg4 2.Re4 Rhe8 3.Rd4 Qxf4 4.Rxf4 Bxf4 5.Qa8+ Kd7 6.Qxb7 a5 7.Rd1+ Bd6 8.Qa6 Re6 9.Qxa5 Ke8 10.g3 Kf8 11.a4 Rb8 12.b3 Rbe8 13.Qa6 Re1+ 14.Rxe1 Rxe1+ 15.Kg2.
"My" stockfish seems to be normal. So the problem may comes from lichess, maybe ^^.
Donterr, I rather think this is a bug because when you end up being mate in 2 forcing moves the computer should not tell you you still have the advantage :/
I used stockfish 6 in Tarrash and set up the same position and the engine gave me this line 1.Qa7 Qg4 2.Re4 Rhe8 3.Rd4 Qxf4 4.Rxf4 Bxf4 5.Qa8+ Kd7 6.Qxb7 a5 7.Rd1+ Bd6 8.Qa6 Re6 9.Qxa5 Ke8 10.g3 Kf8 11.a4 Rb8 12.b3 Rbe8 13.Qa6 Re1+ 14.Rxe1 Rxe1+ 15.Kg2.
"My" stockfish seems to be normal. So the problem may comes from lichess, maybe ^^.
Donterr, I rather think this is a bug because when you end up being mate in 2 forcing moves the computer should not tell you you still have the advantage :/
similar with some theory of why http://en.lichess.org/forum/game-analysis/analysis-is-wrong
This really should have a Q&A entry since it comes up a lot...
Lichess is utilizing several servers to conduct Stockfish analysis, each server generates part of the analysis. And analysis is cut off after a certain time frame, thus limiting calculations. This makes it so that you can get a reasonable analysis on the fly with very little wait. But it also means there are limitations to that analysis and sometimes errors.
So not only is depth an issue, but also the fact that the analysis can be truncated immediately after it finds what it incorrectly thinks is a superior line. If given 2 seconds more, it might see that you are mated and drop back to something it already had determined was best, but since the calculation was truncated, this partially flawed line is what displays. Furthermore, there is always the chance of a hiccup somewhere in the reconstruction of the analysis data from across the servers, or analytical limitations when many people are analyzing games at the same time, putting strain on said servers.
This means that while generally you can trust the evaluation and the first few moves of a proposed alternative in analysis, you shouldn't trust the entire proposed alternative as being accurate. It isn't a "bug" per se, it's just the limitation you get from having fast analysis available without the need to download the PGN and analyze it with your own software.
Lichess is utilizing several servers to conduct Stockfish analysis, each server generates part of the analysis. And analysis is cut off after a certain time frame, thus limiting calculations. This makes it so that you can get a reasonable analysis on the fly with very little wait. But it also means there are limitations to that analysis and sometimes errors.
So not only is depth an issue, but also the fact that the analysis can be truncated immediately after it finds what it incorrectly thinks is a superior line. If given 2 seconds more, it might see that you are mated and drop back to something it already had determined was best, but since the calculation was truncated, this partially flawed line is what displays. Furthermore, there is always the chance of a hiccup somewhere in the reconstruction of the analysis data from across the servers, or analytical limitations when many people are analyzing games at the same time, putting strain on said servers.
This means that while generally you can trust the evaluation and the first few moves of a proposed alternative in analysis, you shouldn't trust the entire proposed alternative as being accurate. It isn't a "bug" per se, it's just the limitation you get from having fast analysis available without the need to download the PGN and analyze it with your own software.
This topic has been archived and can no longer be replied to.