The above position is given in one of GM Artur Yusupov's books. Yusupov has been coached and mentored by GM Mark Dvoretsky, considered by many to be the world's best trainer. In addition, Yusupov, a FIDE Senior Trainer, has long collaborated with Dvoretsky and between them they have helped many successful GMs, including Anand and Leko. Take a minute to think about what you would play.
The position is interesting and it was posted on one of the forums by a player who made the following comments: If you give following position to a chess engine, it probably says the best move is 1.Ne4 or 1.Bxc4. But an expert will (probably) say 1.Bh6, which stops black from castling and the black king remains in the centre. So don't trust engines always!
Out of curiosity I decided to see what some various engines’ top three choices were. After about 3 minutes thinking time, here are the results:
Fire 2.2 xTreme GH x6
After 1.Bh6 0.00
Houdini 1.5 x64
After 1.Bh6 0.00
Stockfish 2.3.1 JA 64bit
After 1.Bh6 0.08
After 1.Bh6 0.01
After 1.Bh6 0.43
Interesting that this is the only engine that selected 1.Bh6 as its first choice.
Critter 1.6a 64–bit
After 1.Bh6 0.08
One 1900 rated player commented: First thought that came into my head was Ne4-Bxc4 and go from there. I like centralized knights. Besides, Bh6 focuses black's play on the queenside, whereas it's more spread out if you don't play Bh6. Not that Bh6 isn't perfectly fine. Apparently he disagrees with Yusupov.
King stuck in the centre is a sudden death in majority of cases. In this particular example Bh6 is also a positional accuracy - king's potential castle position is weakened (dark squares), and in case of fianchetto structure bishop has to be on g7/2, only later can it roam somewhere. This is something chess engines can't see yet. Good comment.
All three moves are good and have up sides and down sides. All three moves should be considered by the expert, and then decided upon thereafter based on taste and experience. An interesting comment. He is correct that all three moves are reasonable, but apparently he would also disagree Yusupov’s advice that you should play the best move…assuming you know what it is. I disagree with that philosophy, but at the same time see his point. If one is better at tactics than strategy or does not understand the strategic implications of the position, then go for a tactical line. Of course here does 1.Bh6, preventing Black from castling and forcing him to keep his King in the center not have tactical implications? He is correct that engines often do not make the best moves when it comes to positions where strategic considerations are important.
What move did I select? First I considered 1.Bxc4 0–0 2.Be3 Rb8 3.Bb3 followed by Qd2, centralizing the Rs and then playing Bh6, but because Black can’t defend the P anyway, finally decided on 1.Bh6 first to prevent Black from castling. But that was without spending a lot of time on the position. After more analysis I might have changed my mind.
After deciding on my move I looked at 1.Bh6 Bf5. This is interesting because now if White continues with the routine 2.Bxc4 Rb8 3.Bb3 then after 3…c4 4.Bxc4 Rxb2 both his N on c3 and c-Pawn are under attack but the position, according to the engines, is equal after 5.Qf3. To me it looks like White is losing control of the game. What White has to play after 1.Bh6 Bf5 is 2.g4! driving back the B and then he can safely take the P on c4. The scary thing is that 2.g4 appears to weaken White’s own King’s position.
After 1.Bh6 Rb8 it seems, analyzing with an engine, that Black has plenty of play on the Q-side, so the truth is, if I had this position in a game on LSS I would have probably selected 1.Bxc4. Given more time in a game played without engine assistance I might have finally decided on 1.Bxc4 anyway. Because Komodo won the contest in selecting 1.Bh6 we will rely on it to tell us that after 1.Bh6 Rb8 play should run 12.Ra2 Be5 3.Bxc4 Rb4 0.16.
What did Yusupov say? Winning the c4-pawn is not important, but stopping Black from castling is worth more than winning a pawn. Unfortunately I do not have Yusupov’s book so can’t say what his analysis is, but it would be interesting to know.
This is just another example of why, even playing with an engine on LSS or ICCF, it’s impossible for us rating challenged players to outplay people who really understand chess.