Sharp play by Black: he intends to maneuver the knight to b5 and possibly close White's only open file at some point with ... Nd4. White must prevent this maneuver if he is to regain the initiative.

Black cannot afford to let the g-pawn fall. 37. ... Nb5; 38. Rxg7, Nxa3; 39. Bh6 Nxc2; RxBf7+ (wins the bishop outright), K-any; 40. Rxb7 followed by 41. Ra7 holding the Black a-pawn.

This renders ... Nb5 unpalatable, and lays the groundwork for a critical break.

The Black queen-side now appears fixed and stable, and White's c-pawn might now appear doomed to backwardness. But White has a surprise in store.

An unpleasant decision for Black. Pretty much any other move, allowing White to exchange on b5, hands White two open files for the rook and an easy end game.

At last, the White bishop comes out of hibernation! White has regained the initiative and his end game prospects are looking good now.

A weak move. Relatively better was ... Be8, to prevent White's next move. However, the further defence of the Black position would have been very difficult against White's combined rook and active bishop on c4.

See note to move 20. ... h6: Black weakened his king-side, and White now invades here decisively.

Again, Black cannot afford to let the g-pawn fall, as this would give White a second passer. The rook now mops up the rest of the Black pieces.

This allows White to pin the knight and then exchange it, leading to a won end game. Black should hang onto this knight as a blockader against White's passed b-pawn, reserving his king to defend the king-side. Therefore, slightly better was ... Nd4 or ...Nc3.