-1

In this position (White to play)

enter image description here
(Picture credits to the chess compass extension that analyzes a chess.com game after it ended)

The current Stockfish evaluation (before White plays) is ~0.0.

But SF would give a -0.8 (+0.8 for Black) if Bg5 takes Nf6 (playing around with SF, the rating goes from -0.9 [picture] to a max at -0.3). Next, SF recommends to take the B, gxf6.

Why would that be?

  • Is it because a B is usually worth more than a N?
  • or in this particular position B had better opportunities rather than take the N?
Déjà vu
  • 605
  • 4
  • 13
  • 2
    Actually, the white king ends up more vulnerable with gxf6, not the black king. The black king doesn't end up in the open because there is no way for white to put a rook in a place to attack it. On the other hand, black can play Kh8 and Rg8 to attack the white pawn at g2. – Alexander Woo Nov 05 '22 at 04:34

1 Answers1

2

You seem to have made some random repetition moves Ra1-Re1 and Ra8-Rb8 before this position, so that Ra1 would be a draw as indicated by your graphic. This makes no sense otherwise, because Ra1 hangs the knight on e2. If not for the previous moves that lead to threefold repetition, the real evaluation of this position is that black is much better (more than -1 advantage, possibly winning) even though Bxf6 is the best move.

Why is Bxf6 the best move? This is an endgame (or late middlegame), and Bxf6 doubles blacks pawns and isolates the h pawn, both of which can become weaknesses for you to attack. Possibly even more important than this is that the black knight has prospects to jump to e4 or d5 and wreak havoc on white's position, while the white bishop is awkwardly placed blocked by white's own pawn on f4. Thus, this is a trade of a bad bishop for a good knight which is a good idea for white.

Nevertheless, white is still worse because black has the bishop pair and more space on the queenside. White's pieces are placed somewhat awkwardly, especially the b3 knight which is trapped by black's pawns and unable to advance forward. The knight on e2 blocks white's control of the e file, but cannot easily be moved without hanging the f4 pawn. The piece coordination is nonexistent in white's camp, but perfectly harmonious for black.

YiFan
  • 616
  • 5
  • 12
  • It's not how it happened. I tried a few variations to see how the evaluation goes. The evaluation was ~0.0 *before* white moves, assuming white plays the best move ; that's how SF works (after X plays, rating can only stay put or go down for X - unless X finds a better move than SF, or SF does a 1 move deeper analysis and find a different rate). Among the relevant possible moves, SF gave +/-0 for Ra1, and ~ -0.5 for Bxf6, which would indicate Bxf6 is not the best move. Maybe trying SF on a different tool (chesscompass could be buggy) would give a different result? I'll try when time allows it. – Déjà vu Nov 06 '22 at 05:32
  • 1
    @Déjàvu Bxf6 is definitely the best move, and the position is definitely worse for white. The real evaluation for Ra1 is something like -10 because it hangs a piece. This is not how SF works; if the best move in the position is still worse for white, it is not going to pretend like the position is equal. If chesscompass is claiming that this position is equal and Ra1 is the best move, but you didn't play the Ra1 repetition, then that means the chesscompass engine is hopelessly flawed as it doesn't even see a free hanging piece on e2 and you should stop using it. – YiFan Nov 06 '22 at 05:41
  • Of course it's not - and it's now what I said :) Say an eval of a position (all pieces) P is E, white to play, then unless white plays the top SF move, E will (usually) go +/- down for white, because SF assumes the moves to come match its own best eval (then depth matter), rest is worse (same for black, thus the bar moves up and down). Ok, so since chesscompass eval and yours seem to differ a lot, if Bxf6 is indeed the best move, the question is moot. – Déjà vu Nov 06 '22 at 06:17
  • 1
    @Déjàvu I'm not interested in debating whether the numerical SF evaluation goes up or down, that's totally irrelevant to the chess and irrelevant to the question you asked. I answered your question based on the chess, and I'm not sure why you're trying to argue with me based on some kind of flawed analysis by an engine that can't even spot a piece hanging in one move. (Or, more likely, that you really did make the repetition and are just not admitting it for some reason.) – YiFan Nov 06 '22 at 06:30
  • Your answer suggests there was a repetition draw, which was not the case, and anyway, if it was a draw the UI wouldn't suggest further moves, that motivated the SF algorithm explanation. Thank you for your position analysis, but the question asks specifically why the SF eval tells Bxf6 is a bad move, while it looked good at first glance. Again, this has to be tested on another interface. And, again, if the eval was wrong initially, the question is moot. – Déjà vu Nov 06 '22 at 06:54
  • 1
    I did not say there was a repetition draw, only that Ra1 in this position would be a draw by repetition. The engine will suggest further lines in this case, what it's telling you is that Ra1 would be a draw by repetition which is why it's 0.0. There is no other sensible explanation for Ra1 being equal. In this case, not taking the draw as white is bad because white is worse. – YiFan Nov 06 '22 at 07:07
  • 1
    You don't need to test on another interface to realise that Ra1 is losing if not for the repetition. You see this by noticing that the knight on e2 is hanging. If you insist: https://lichess.org/analysis/r3r1k1/p4ppp/1pbb1n2/2p3B1/5P2/1NP4P/PP2N1P1/4RRK1_w_-_-_0_1 You can make the move Ra1 yourself and see that it is almost -10. – YiFan Nov 06 '22 at 07:09
  • 1
    Thanks, will look at it. – Déjà vu Nov 06 '22 at 07:10