The extremely sharp Dragon Sicilian position shown in the first diagram arose in the game Palac-Lalic, Pula (Croatian Team Ch.) 2000. It is White to move. Both sides appear to be attacking, and it is not easy to say whose chances are better. The immediate question is whether the knight on c3 must be defended. More precisely, can White ignore it and play 1.Bxh6 right away?
In the game, Palac went for safety with 1.Rd3?! This could and should have allowed Black to barnacle with 1...g5!, but Lalic instead played the mistaken 1...h5? After 2.gxh5 Nxh5 3.Qg2! Black's draughty king position gave White a strong initiative and his attack eventually prevailed.
Some time later this game came under the scrutiny of the noted Dragon theorist Mikhail Golubev, a grandmaster from the Ukraine. He recommended the following continuation of White's attack: 1.Bxh6! Rxc3 2.Bxg7 Kxg7 3.Qh6+ Kf7 4.g5 Ne8 5.Rh4! e5 6.Qh7+ Kf8 7.Rdh1! Here Golubev writes that Black is helpless against the threatened 8.Qxg6 exd4 9.Rh8+ Ke7 10.Rxe8+ and his analysis ends here.
After 7.Rdh1 Black cannot play 7...exd4 because of 8.Rf4+ and wins, so he must find a different move. If we give him the superfluous 7...a5, Golubev's line can be continued as follows: 10...Bxe8 11.Rh7+ Kd8 12.Qxd6+ Kc8 13.Qe6+ Kd8 14.Rxb7 Rxb7 15.Qf6+ Kc8 16.g6! Here the legendary Mikhail Tal would certainly adjudge this as winning for White, and very few would disagree with him.
In passing let us observe that instead of 6...Kf8 Black cannot block the queen check with 6...Ng7 on account of 7.Rh6! Rg8 8.Qxg6+ Ke7 9.Qxd6+ Kd8 10.Qxe5 and with four pawns and a raging attack for the piece White must be easily winning. This idea will resurface in a later note.
Of course there must be something stronger for Black than the superfluous 7...a5. At the very least one could look at 7...Qc7, which gives White almost the same variation with the key difference that the pawn on d6 is now defended. Now After 8.Qxg6 exd4 9.Rh8+ Ke7 10.Rxe8+ Bxe8 11.Rh7+ Kd8 12.Rxb7 Kxb7 13.g6 Black can give back a piece with 13...Bxg6 and fight on with two rooks against White's queen. Whether he would be successful is another matter, but the point is that this is not clearly a forced win for White.
Let us peel back to the position after 4...Ne8 (second diagram) and make White play 5.Qh7+! instead of 5.Rh4. With the square e7 still blocked by his pawn, Black cannot play 5...Kf8 since 6.Qh8+ followed by 7.Rg7+ delivers a quick mate, so Black must reply to 5.Qh7+ with 5...Ng7. Unfortunately this allows White to switch back to the variation analysed above with 6.Rh4! e5 7.Rh6! As before Black has nothing better than 7...Rg8 8.Qxg6+ Ke7 9.Qxd6+ Kd8 10.Qxe5 and a win for White is just a matter of time.
Varying the order of one's moves and gauging the effect is an important part of attacking technique. I wish I could learn to do this properly during actual play!