<Up to here everything had been played quite a few times. It would seem
that 15... b4 was the older, and probably better choice, as played, in,
inter alia, Movsesian-Orsag, Czechoslovakia 96. 15...Bb7 was apparently
an improvement (?!) recommended by Hellers (or was it Hector?? I
really must get a new memory...) (See Diagram)
The start of the refutation. Previously the timid 16.Re1 had been seen in
Degerman-Cvitan, Vienna 91, and Hector-Vogt, Taastrup 90. 16.Rg1
takes aim at the apparently secure g7 point.
16...b4 17.Nd5!! (See Diagram)
The lovely point. White sacrifices a piece for a tempo.
17...ed 18.Rdg3!!! (See Diagram)
To double on a closed file!! A unique idea. The key is that the g-file will
be blown open by either the crude g6, or the more elegant Qh6. Black is
totally defenceless. Incidentally, the immediate 18.Qh6 fails to 18...gh
19.gh Bg5. But with another rook on the g-file, there is no way out. The
lines are not difficult anymore. My restaurant card notes give 18...g6
19.Qh6 f6 20.gf Rf6 21.Rg6 hg 22.Rg6 Kf7 23.Rg7 as the longest
variation. After that there is a soup stain... It is only the concept of
white's last couple of moves that is so striking. I wonder if the top
computer programmes could find it?? Cvitan now sank into a huge think,
but there is no escape...
18...Qc7 19.Qh6 Qc2
The only way to make space for his king.
20.Kc2 Rfc8 21.Kd2 gh 22.gh Bg5 23.Rg5 Kf8 24.ed Ke7 25.Rf5 Rc4
26.Kd3 Rac8 27.Rg7 1-0>