notyetagm: N Dzagnidze vs S Guramishvili, 2013
Of course <HOUDINI 3> immediately calculates that the speculative sacrifice 16 ... ♗d6xg3? is unsound, giving it an eval of <2.10 >.
Also note how <HOUDINI 3> also immediately sees that taking the other way (17 h2x♗g3?) is losing, giving that move an eval of <-2.19 >.
I wonder if Guramshivili (Black) considered the enormous difference between these seemingly nearly identical recaptures. Probably not. This is a point that I have seen NM Jorge Sammur-Hasbun make over and over again in analysis sessions: it can make a *huge* difference whether you play 17 f2x♗g3! or 17 h2x♗g3? , even though there seems very little difference between them at first glance. Only deep <CALCULATION> reveals the truth.
<CALCULATE LIKE A COMPUTER!> Dzagnidze playing White did, and she won easily as a result.
Computerkibitz: Houdini 3 Pro x64 Depth: 20 with 302.619 kN in 110s
<2.10 <<<17.fxg3>>> Bxg2 18.Kxg2 Nd6 19.h3 Qh5 20.Qd3 Nc4 21.Rxc4 dxc4 22.Qxc4 Ne7 23.e4 b5 24.Qc5 Qg6 25.Nc3 Qd6 26.Bc1 Rad8>
<-2.19 <<<17.hxg3>>>> Nxg3 18.fxg3 Qxg3 19.Nh4 Qxh4 20.Bxh3 Qxh3 21.Rc3 Qh4 22.Kg2 Nxd4 23.Rh3 Qe4+ 24.Kh2 Nxe2 25.Nc5 Qf4+ 26.Kg2 d4
-13.29 17.Bxh3 Bxf2+ 18.Kh1 Qxh3 19.Rg1 Bxg1 20.Qxg1 Re6 21.Nd2 Nxd2 22.Nc3 N