Last week’s puzzle was taken from Aussie GM David Smerdon’s terrific new book The Complete Chess Swindler. My review will be published on British Chess News shortly.

This is Shirov-Kramnik (Groningen 1993). White’s attack has misfired and he didn’t fancy 18. Qh4 Nxg3 19. Bxe7 Nxf1 20. Bxd8 Qxe5 21. Bf1 Qe3+ 22. Kb1 Bc6 when Black stands better.

Instead he chose the queen sacrifice Bxh6, the move I asked you to evaluate last week.

The game concluded 18… Nxf4 19. Bxg7+ Kh7 20. Rxf4 Rg8 21. Rfg4 Rxg7 22. Rxg7+ Kh6 23. Rg8 Kh7 24. R8g7+ with a draw, but Kramnik had missed something.

He could have returned the queen with either 20… Qxc3 or 21… Qxc3, when taking with the pawn allows Ba3+, mating, while taking with the rook would have left him a piece ahead.

So the correct answer to my question is that Bxh6, even though it worked out over the board, should lose with best play. Whether or not it was the best practical chance is another matter. At what point, if at all, did Shirov see Qxc3?