Here is the post I have just made about it on Chessworld.net:-

Alexander Kotov in his classic renowned chess book "Think like a Grandmaster" advocoated looking at the way machines think - especially in relation to tactical variations.

Curiously on the rec.games.chess.misc newsgroup recently, the very first position that Kotov presents in his famous book is question for the variations examined- in particular the deliniation of candidate moves.

Kotov gives the following candidate moves:-

a) Bxh6
b) Nxg6
c) Ng4 with idea of Nxh6+

Nxf7 is curiously missing - and this actually goes against a fundamental principle emphasised by the book of allowing a "brainstorming" process of candidate moves- not rejecting moves too early.

A recent rec.games.chess.misc reader Henri H. Arsenault, has run this position through Fritz to apparently test their hunch that Nxf7 is the strongest and most obvious first move, to undermine the e6 square and maximise the impact of the Rook on e1 and Bishop on b3.

Variations given include:-

1.Ne5xf7 Qf6xf7
(...Kg8xf7 2.Re1xe6 and if Black does not take the Rook with the Queen, the discovered check will be fatal)
(...Bb8xg3 2. Nf7xh6+ Kg8-f8 3. h2xg3 and White is a piece ahead with a strong attack)
(Ng6-f4 2. Bb2xf4 Qf6xf4 3. Qh5-g6 and Black will have to give up the Queen to avoid checkmate)
2. Bb3xe6 wins the queen

Not only this but one of the key candidate moves given by Kotov - namely Ng4 actually gives black the advantage after Qxd4

e.g. Ng4xh6+ Kg8-f8 that leaves Black with a slight advantage.

This is quite astonishing given the fact this is the very first position of a classic book, which claims in a recurrent theme that we can improve our tactical ability through trying to think systematically like a machine. I expect that many of the old chess books will have their variations crunched up and refuted, and better more decisive lines found by today's engines!