Rybka 1 - Fischer / Huebner / Kasparov 0

Continuing with Fischer - Gligoric, CT 1959, in the diagrammed position Fischer played 23.Qd3. Annotating the game, he assigned his move '!' and wrote,

The more obvious 23.f5 looks good, but Black still has defensive resources with 23...exf5 24.Rxh5 gxh5 25.Nf4 Rxc3 26.bxc3 Rxc3 27.Qxh5 Rxb3+ 28.cxb3 Qe3 etc.

According to Kasparov, Huebner gave Fischer's move '?', because of the variation

23.f5 exf5 24.Rxh5 gxh5 25.Qxh5 [instead of 24.Rxh5] Be6 26.Bxe6 fxe6 27.Qg6 Qc7 28.Rh1 Qe7 29.Nf4 Qe8 30.Qh7+ Kf8 31.Rh6 Bxh6 32.Qxh6+ Ke7 33.Qxe6+, 'when the Queen and Knight make short work of the Black King'.

Kasparov assigned Fischer's move '!?', because of

23.f5 exf5 24.Rxh5 gxh5 25.Qxh5 Be6 26.Bxe6 fxe6 27.Qg6 Re5! [instead of 27...Qc7] 28.Nf4 d5

with analysis of 29.Nd3, 29.Nh5, 29.Rh1, and 29.Ka1.

1959 Candidates Tournament (round 4)
Gligoric, Svetozar

Fischer, Robert
(After 22...Rc6-c5)
[FEN "2r3k1/pp1b1pb1/1q1pp1p1/2r3Pn/4PP2/1BN2Q2/PPP1N3/1K1R3R w - - 0 23"]

Fischer and Kasparov, joined this time by Gligoric, then settled into analysis of 23...Bxc3, which everyone assigns a '?'. While studying the analysis, which hinges on the sacrifice Rxh5, I wondered why Fischer couldn't play it on his 23rd move. I plugged 23.Rxh5 into my software, which quickly came up with 23...gxh5 24.Qxh5 Be8 25.f5 exf5 26.g6 Qc7 27.Nd4, when White has a nice attack.

I then wondered whether anyone on Chessgames.com (see link on the original post) had pointed this out. Indeed:

Fischer missed the winning 23. Rxh5!!. After 47 years a piece of chess software called Rybka, the strongest commercial chess software today, spotted this move overlooked by Fischer, all GM commentators, including Kasparov, who only gave the alternative 23. f5!?.

The analysis, given by another kibitzer, goes 23.Rxh5 gxh5 24.Qxh5 Be8 25.f5 exf5 26.Ng3 [instead of 26.g6] 26...fxe4 27.Ngxe4 etc., or 26.Nf4 etc. See Chessgames.com for the moves after 'etc.'

This Rybka contraption is frightening!