Why Bxa2 is bad according to Silman?

Question

[FEN "2nr1r2/1pp1q1pk/2n1bp1p/p3p3/4P2N/2P1N1P1/PP2QPBP/3R1RK1 b - - 0 1"]

1... Bxa2 2. Bh3 Nb6 (2... Nd6 3. Rxd6 cxd6 4. Bf5+ Kg8 5. Ng6 Qf7 6. Nxf8 Qxf8)

I found the above position in How to Reassess Your Chess by Jeremy Silman. He said that after 1...Bxa2 2. Bh3 Black is suddenly dead meat. But according to Stockfish, 1...Bxa2 is the best move.

Why did Silman think Bxa2 is bad?

Silman wrote in the starting page that he used Fritz and Deep Blue. They are not better than Stockfish as everyone knows. But Silman also added that he got help of some IM and GM. Also, he analyzed the positions.

Instead his continuation was 1...Nb6 2. Nd5 Bxd5 (Stockfish suggests Qf7. As a human, I also don't like Knight on d5 so I will also consider taking it with the bishop since knight take d5 is blunder cause exd5 forks bishop and knight) 3. exd5

Answer 1

I agree with Hauptideal's general answer, but for this specific position perhaps I can add a suggestion why 1...Bxa2 might not be very human even if the engine does enjoy the position afterwards.

The problem here is the f5 square. Both of white's knights are staring at that square, and if a knight does land there it does so with a tempo on the black queen, and very near the black king too. You could imagine that in some worst case scenario, if the black queen wanders off to the queenside, say, white might be able to play moves like Qh4 at some point and black would have to worry about king safety. Even if not, the fact that there is a knight on f5 that is more or less impossible to remove (even if you try to trade with ...Nd6, the other white knight takes its place) is not very nice from a human perspective and white gets more than enough compensation from the pawn, or at least so it would seem to a human playing this game.

If 1...Bxa2, the point of Bh3 is not just to control the light-squared diagonal but also more concretely the f5 square. There are ideas of not just Nf5 now, but also Bf5. To help to understand this position consider a concrete line where I will play human moves for black and engine responses for white: 1...Bxa2 2.Bh3 Nd6 3. Rxd6! Qxd6 4. Rde1 Qe7 5. Bf5+ Kg8 6. Nd5 Bxd5 7. exd5 Nb8 8. Be6+ and black is lost.

[FEN "2nr1r2/1pp1q1pk/2n1bp1p/p3p3/4P2N/2P1N1P1/PP2QPBP/3R1RK1 b - - 0 1"]

1... Bxa2 2. Bh3 Nd6 3. Rxd6 Qxd6 4. Rd1 Qe7 5. Bf5+ Kg8 6. Nd5 Bxd5 7. exd5 Nb8 8. Be6+

If you ask the engine, turns out the critical mistake was 6...Bxd5, but this doesn't seem like an outrageous move to a human because the black bishop is stranded on a2 and this not just solves the problem of the off-side bishop but also eliminates a seemingly powerful white knight. But concretely, black loses due to weaknesses around his king. Even if black does not fall victim to this blunder, if you look at the lines, there is often only one move that maintains equality or very slight advantage for black (less than -0.5, let's say) and every other move is a huge advantage for white. This is why the position is tough from the perspective of human intuition and strategy, even at GM level, despite the fact that Stockfish seems unconcerned about grabbing the pawn.

Answer 2

In every classic book, you'll have lines that modern engines do not like. I stumbled on the very position when I read the book, and I remember the phrase "dead meat" being extremely cryptic to me. However, I found, that Silman's books are much better than, say, Chernev's Logical Chess Move by Move (which I stopped reading precisely because of the bad objective quality).

It is very good that you check everything with an engine, that you do not understand. However, even if a modern engine considers a variation better than Silman, it is still very valuable to read and understand Silman's annotations, because you'll learn something about practical "human" moves and how to generally think about positions, even if concretely, a certain other variation might work slightly better. Often, these better variations are hard to assess for humans correctly, because a lot of complications are to be considered or a great calculation depth is needed to see the point. In short, they are impractical, and it's hard to deduct a general strategic pattern/rule from these moves. The goal of "How to reassess your chess" is not to make stockfish-like calculation machines out of you, but to teach you strategic concepts, and his moves are always "good enough" for practical play and learning strategic ideas. Especially in a book about strategy, there will be alternatives and multiple possible avenues, but he chose one specific line to demonstrate an idea to you. Of course, you may not have to maneuver your knight to that juicy outpost, for example, but it's pretty good to do it and he can prove his point that you sometimes can and should invest the time to get your pieces to good squares. Yes, there may be another variation the computer might play, but you wouldn't have learned the concept.
In a few years, a new chess engine might laugh at your Stockfish 15, just as Stockfish 15 now laughs at the engines Silman used to write his books. The strategic principles, however, stay. When you find an example, where Silman is very wrong objectively, you should of course skip this particular example or not take it too gravely (in the example, the move is so impractical for Black that even an International Master would refrain from doing it (even if objectively, the ?? may be wrong, it may be correct practically..). This is the exception in this book, anyhow.

Should I throw the PDF right now?

Yes, but only for you to read the hard copy. Please buy the book if you can afford it. There are no legitimate (scanned) PDFs from Silman's works. But I would keep reading because you can still learn a lot from Silman.