What a great endgame to analyze!
Where should I start with???
What was the point of Shirov's 47…Bh3?
We know that 3 healthy pawns win against a bishop in endgame.
Tablebases show the following endgame as won for Black , if it is his turn to move, by playing
Kf5 ( the only winning move!) :
[fen "8/8/4kp2/3p4/p7/2B5/8/6K1 b - - 0 1"]
Black will simply stretch White's pieces and win-it is a standard motif and I believe that any strong FIDE master ( or higher of course ) can win this endgame with ease.
Looking at the position after
...Bh3!! we can see that either side's king side pawns play no significant role in the game, so we have the above endgame, with "spectators" on the king side.
Since we know that the endgame is won for Black with inactive king side pawns it is not hard for us to find
...Bh3!!, as Black will implement the same winning plan like in the above endgame.
Still, I doubt Shirov knew the above endgame "by heart", so he had to calculate a lot to confirm that the "pawns vs bishop" endgame is won, thus "a tip of the hat" from me.
I personally would never even think of playing such a move, to be honest.
Can anybody give an explanation of why it is the best move?
Since both sides king side pawns are spectators, it is really the above mentioned "pawns vs bishop" endgame. With 3 passed pawns Black easily overloads White's pieces, and forces a fast win.
I personally thought to play
...Be4, but since I have 2 passed pawns, and my King is somewhat blocked, I had to work harder to force a win.
I played the below game against engine. Each time I finish playing one variation, I would set another and play it. Also I have added my own analysis of the moves computer refused to play. That way all the relevant moves would be covered. The key was for Black king going to the queen side, overloading White king and bishop by creating protected
d3 pawns, after which Black reroutes king to the king side and grabs the pawns with decisive advantage. It is clear why this is longer and harder way of playing, then more straightforward way Shirov picked. Here are the lines:
[FEN "8/8/4kpp1/3p1b2/p6P/2B5/6P1/6K1 b - - 0 0"]
1... Be4 2.Kf2 (2. g4 Bf3 3.g5 fxg5 4.hxg5 Kf5 5.Kf2 Bh5 6.Bf6 Ke4 7.Ke1 Kd3 8.Be7 d4 9.Bc5 Kc4 10.Be7 Kc3 11.Ba3 d3 12.Bc1 Kc2-+) (2.g3 f5 3.Kf2 Kd6 4.Bd4 Kc6 5.Ke3 Kb5 6.Kd2 Kc4 7.Bb2 Kb3 8.Kc1 a3 9.Ba1 Kc4 10.Kd2 d4 11.Ke2 Bg2 12.Kd2 Kd5 13.Ke2 Ke4 14.Kd2 Bf1 15.Kc2 Bc4 16.Kd2 Kf3 17.Bxd4 Kxg3 18.Ke3 f4+ 19.Kd2 a2 20.Be5 a1=Q 21.Bxa1 Kxh4-+) 2...f5 3.g3 Kd6 4.Bb4+ Kc6 5.Ke1 Kb5 6.Bf8 Kc4 7.Kd1 d4 8.Kc1 d3 9.Kd1 Kb3 10.Bg7 a3 11.Bd4 a2 12.Kd2 Kc4 13.Ba1 Bc6 14.Bb2 Bb5 15.Bc3 Kd5 16.Ke3 Bc4 17.Ba1 Ke6 18.Kd2 Kf7 19.Kc1 Kg8 20.Bd4 Kh7 21.Kd2 (21.Kb2 Kh6 22.Be3+ Kh5 23.Ka1 Kg4 24.Bf4 Kf3 25.Kb2 Ke2-+) 21...Kh6 22. Ba1 Kh5 23.Ke3 Kg4 24.Be5 Bb5 25.Kd2 Kf3 26.Bc3 (26.Kd1 Ke3-+) (26.Kc1 Ke2-+) 26...Kxg3-+
It took me a lot of moves to win this endgame, and the pitfalls where you can squander your advantage are numerous-White indeed can offer some resistance.
Since king side pawns are irrelevant, we can roughly conclude that Shirov's way wins in 26 moves, according to tablebase, without giving White any chance to offer any resistance.
Hopefully this answer helped. If you have further questions leave a comment.