lichess.org

Why is the Engine move in the Italian Game vs Two Knights Defense different from it's color mirror?

lichess.org/study/CEcyy9Dx

In the position that defines the Italian Game vs Two Knights Defense the Engine move is for white to go for the knight attack, but in the position that has perfect mirror color symmetry to this position the engine move is to move a pawn despite the fact that each position can be transformed to the other position by reversing both the dimensions and the colors. Also the evaluation for the mirror color position is different from the opposite of the evaluation of the original position.

lichess.org/MO6fKxaM/black

In this game I played for instance the knight attack was labeled an inaccuracy despite the fact that it is the engine move in the Italian Game vs Two Knights Defense, that the position on move 4 with black to play is mirror color symmetric to.

Do not believe engine comments on openings. Engines are weak in openings without their opening book derived from human practice. The opening with 32 men is far too complex for the engine to get a grap of it. Any relevant events lie behind its calculation horizon. It is only after some moves have been played and a few men have been exchanged that the real strength of the engine kicks in.

@acgusta2 Stockfish has no code at all meant to evaluate things differently depending on the color, however the pawns PSQT are assymetric, which means that a vertical symmetry transforming the h file into the a file, the g file into the b file, etc., will lead to subtly different evaluations.

@tpr Humans are very weak in the openings without their memorized opening moves, too, significantly worse than engine at that actually. Almost all Chess960 game between top masters are filled with big mistakes in the first ten moves. You don't need to use human games to compute an engine opening book, either.

That said, Stockfish on low depth is definitely not to be trusted too much, especially in openings.

@Alayan this is truly 100 percent symmetric situation e3,e4 just loses a move and is basicallly same as pass on firts move. THere is absolutely no differences. exept that board is mirrored on vertical plane but thats like haveing switcked kings and queens around. Changes nothing. Truly odd that it gets different evaluation as evaluation is color agnostic.

This is an „alter Hut“ (old hat). Engines have different evaluations for mirrored positions, sometimes they give a small plus merely for having the move and so on.

Well, this knight sacrifice is rather doubtful. After 1.e4 e5 2.Nf3 Nc6 3.Bc4 Nf6 4.Ng5 d5 5.ed N:d5 6.N:f7?! K:f7 7.Qf3+ Ke6 8.Nc3 Nb4! 9.a3 N:c2+ 10.Kd1 N:a1 11.N:d5 White was long thought to be winning, but discovery of 11...Qh4!! completely overturned the evaluation. Hence, people are trying to deviate on 9th move, but without real success - in best case, White may achieve dynamic equality. Therefore, the logical choice, instead of sacrifice, is 6.d4! with great perspectives for White.

Nowadays, I generally employ Caro-Kann, but some years ago I really liked to defend against this knight sacrifice.

That's just not true. Of course you can play 6.d4 but 6.Nxf7 is still the main move these days. I don't think anybody thinks black is better in the 6.Nxf7 lines. But we just know now that it isn't necessarily losing with ideal play from black. F.ex: lichess.org/LKNxN8mF

Well, I don't deny that 8...Ne7 might also be holdable with perfect play. Personally, I believe that 8...Nb4 is easier to play, much more fun and probably plainly stronger, but who am I to criticise Agdestein's choice, even if he eventually lost this game?

@Otienimous Ng5 does not yet still mean that one will sac the knight. And fact there is defence which work on top level does not mean it works for you. Nf7 sac works mostly because all B moves are hardet to find thatnW. And this has nothing to do with why evaluation differs on color reverse

Mayre mirrored image gets diffetent value due control of square values engines employ? thos might be non-symmetric

Reconnecting