It is understandable that White should try to bring his Q to the defense of the K-side and to vacate the d-file for a rook. But this was White's last chance to prevent ... Nd5-f4 by exchanging either Nxd5 or Bxd5.

I consider White's position to be untenable here. Black has now amassed enough material to crack open the White bastion by exchanging a piece for two pawns on h3. As a general rule, strategy books say that an attack on the flank will not succeed unless one has control of the center. This is an exception. Black cannot really be said to control the center. The crucial thing is, White's center is static; his e-pawn, although on the Black side of the board, is immobile, and the ostensible weaknesses in the Black camp--d6 and f6--cannot be exploited by White pieces. Black, on the other hand, threatens to exchange off White's K-side defenders, advance the g and h pawns, and penetrate at some opportune moment with Qg5 or Qh4. Black's control of f4 proves to be the crucial launching pad in the coming attack.

This puts the White Q on the same diagonal as the Black dark-squared B, offering Black a tactical shot.

Black easily regains the piece, but, more importantly, the White K is now vulnerable.

Alternative: 20. Ng5, QxNg5; 21. Qd3, h4; 22. d5, g3; 23. dxe, gxf++; 24. Kxf, Qg2+; 25. Ke1, Qg1+; 26. Ke2, Bg4+; and the White Q must interpose to avert mate.

When attacking a king with combined queen and rook, it is generally better to lead with the rook and follow with the queen. Hence ... Rg8+ here and not ... Qg5+ (which would be easily parried by 22. Qg3).

F4 proves to be White's undoing. White resigned here, since 24. Kh1 (forced), QxQ#. 23.Ne2 (to prevent ... Bf4) fails to 23. ... Bg2+ and mate soon after. An example of a flank attack which succeeded, despite not having control of the center.