Black’s position looks critical, as he lacks space and his pieces are uncoordinated. White threatens the break e5-e6, and also wants to play Bg4-f3, with the threat of d4-d5. But there is a brilliant way to change the contours of the game:

The e5-pawn is blocked, and the knight gets a path to d5 via e7.

This is playing on the wrong flank, but even after 26.Bxe6 Qxe6 (weaker, by analogy with the game, is 26...fxe6, because of, say 27.Rf1 Ne7 28.Rg3, with the intention of including the h-pawn in the attack) 27.Rg3 Ne7 28.h4 Nd5, Black obtains positional compensation for the exchange, although not fully adequate.

Even in the case where, after the exchange of the b2- and a6-pawns, White does not lose the d4-pawn, Black’s positional compensation will be sufficient, so a draw was agreed.

It would appear that the position is roughly equal: the pawns are equal, and the opposite-coloured bishops enhance the drawing tendencies. But these considerations would only be true if queens were not on the board. All the while they are, the possibility exists of a powerful attack on the white king.

Black finds a way to break up White’s pawn chain and get at the white king.

White will not manage to get his attack on the enemy king going. It was essential to switch to defence with 29.Kf3 gxf4 30.exf4, although here too, Black has a serious initiative after 30...Qd4 31.Qd2 Bc5.

29...gxf4+ 30.exf4 Qd4!

Now, the black bishop will join in and White’s position is hardly defensible.

White resigned.

If White plays 26.a3, Black gets control of the a-file, whilst after 26.b5, Black gets the c5-square. Exploiting the more active position of his pieces, White found a way to break up the black pawn chain: