http://en.lichess.org/OdVwv43z/black#68It seems like white's only move, but the analysis shows a big shift, even though it's not listed as a mistake. Or why is the previous move such a blunder?
I cannot think in anything but a bug... =/
It's not a bug it's different evaluation at certain depth. If the engine analyzed it longer, it would've found 35. ... Ke6 is a huge blunder. The right move was 35. ... f4! not letting your opponent play f4 himself which in the game cramped your pawns and lead to a draw.
i hate to play in games like this, as DVRazor says the right move is f4. Actually, it looks like the ONLY winning move. Along with continued perfect play. I suppose there are tricks to calculating these pawn moves, but I'll leave it to the GM's
White could sure use an A pawn right about now!
It is that computer was thinking that it was -3.8 and it was close to discovering -1.8 but didn't think long enought and after white's move computer realised that it's actually -3.8
That looks like the computer trying to correct itself during the analysis. Black is winning but the fact that White has those two passed pawns makes things a little complicated. I actually had a game recently where I was Black, the score was -10, I played a move the computer decided I win by force and changed it's score to -#9.
*(From -10 to -#9). This was after Black's move.
It is just a symptom of first to last move analise. If it was to go from last to first move it would solve these things plus having a higher quality analise.