7
[FEN ""]
[startply "16"]
[StartFlipped "1"]

1. e4 d6 2. Bc4 Nf6 3. Nc3 g6 4. d3 Bg7 5. Nf3 Nc6 6. O-O e5 7. Ng5 O-O 8. Bxf7+ Rxf7 9. Qf3

Why does the engine go from -1.5 to -8.5 after the queen move? There isn't an opportunity to trap the queen yet and the black is only two points ahead in material. My guess is that the queen walks into a discovered attack by the rook and the queen can be immediately attacked with Nd4. I played Rf8 afterwards to keep my rook on the f file. The top engine line is just to move the queen back to where it came from and then the knight is kicked out too.

Rewan Demontay
  • 17,514
  • 4
  • 67
  • 113
qwr
  • 3,674
  • 13
  • 39
  • 2
    Why didn't you take the rook? And It's losing for Nd4 and Bg4. – Billy Istiak Dec 25 '22 at 02:37
  • 1
    @BillyIstiak I played as black (idk how to flip the PGN) I don't know why White didn't take back – qwr Dec 25 '22 at 02:40
  • @qwr What engine at what depth is returning -8.5? I'm not getting that unless I run it at, like, depth 3. – D M Dec 25 '22 at 13:06
  • 1
    black is only two points ahead in material Engines are waaaay past the time when their evaluations reflected the material on the board. Plus, these days most engines index their evaluations to chance of victory. – Allure Dec 25 '22 at 13:48
  • 1
    @DM lichess's stockfish 14+ NNUE at depth 26 https://lichess.org/analysis/standard/r1bq2k1/ppp2rbp/2np1np1/4p1N1/4P3/2NP1Q2/PPP2PPP/R1B2RK1_b_-_-_1_9 – qwr Dec 25 '22 at 13:51
  • @Allure I've found Stockfish will report somewhat more than material imbalance when the position is otherwise equal. So a blundered bishop or knight will report say -5. – qwr Dec 25 '22 at 13:52
  • 1
    @qwr Stockfish no longer uses material imbalance. +1 is not the same as "extra pawn with no compensation". +1 now means Stockfish thinks White has a 50% chance of winning. See https://github.com/official-stockfish/Stockfish/commit/758f9c9350abee36a5865ec701560db8ea62004d. – Allure Dec 25 '22 at 14:01
  • @Allure the engine I used is SF14 NNUE – qwr Dec 25 '22 at 14:07
  • 2
    @Allure That's good to know, but that's in the notes for version 15.1, which is a higher version that what was used here, so it probably wouldn't apply here. – D M Dec 25 '22 at 14:07
  • @DM it still applies, because ever since NNUE Stockfish's evaluation (usually, outside of the few positions that use classical eval) depends on win percentage, and Stockfish adopted NNUE in version 12. – Allure Dec 25 '22 at 14:27
  • @Allure Stockfish used to think it did, according to an old answer of mine, although my source seems to have disappeared. https://chess.stackexchange.com/questions/17196/how-are-the-scores-decided-in-chess-correct/17204#17204 – D M Dec 25 '22 at 15:14
  • @allure I just remembered that the wayback machine is a thing. https://web.archive.org/web/20180424122041/http://support.stockfishchess.org/discussions/questions/1303-stockfish-evaluation-numbers-and-their-meaning – D M Dec 25 '22 at 16:16
  • @DM I take back what I wrote - it seems during the HCE era, Stockfish evals indeed were indexed by the value of a pawn, although they used the endgame value of pawns and not the middlegame one. – Allure Dec 26 '22 at 10:36

2 Answers2

9

White is a piece down for just a pawn in this position. Taking the rook would give them a reasonably balanced position, but by refusing to take they get a totally lost one.

There are also some ideas with Bg4 and Nd4 that speculate with trapping the queen or taking on c2 (potentially forking the rook and queen if she moves to the wrong square).

I wouldn't pay too much attention to the specific number. Whether it's -3 or -7 it doesn't matter much. The fact is that White should have taken the rook and any other move is a blunder.

David
  • 16,275
  • 26
  • 61
  • 1
    Another detail coming out bad for White: The knight g5 can't safely retreat, on f3 it will be pinned (the Q only has d1 as safe field) and on h3 exchanged. Thus White, in addition to material woes, also has a trashed castling position. – Hauke Reddmann Dec 25 '22 at 20:36
5

It seems like white played Bxf7, and after Rxf7, instead of taking the rook with Nxf7, white played Qf3. Now black -among others- can play the simple Rf8 and are a piece up for a pawn.

The fact that the advantage is so large just depends on the engine and its version. For example, from Stockfish 15.1 evaluations are not that extreme (in my laptop it's "only" -5).

In any case, white position goes from -1 to completely lost not because Qf3 puts the queen in danger or anything like that, but just because they didn't play Nxf7 (winning a rook for a knight), sort of balancing the material on the board.

Allure
  • 26,534
  • 1
  • 68
  • 146
emdio
  • 2,094
  • 13
  • 22