1.d4 Nf6 2.c4 g6 3.Nc3 Bg7 4.e4 d6 5.f3 0–0 6.Be3 Nc6 7.Qd2 a6 8.Nge2 Re8 9.h4 h5 10.0–0–0 b5 11.Nd5 bxc4 12.Nxf6+ Bxf6 13.g4 hxg4 14.h5 g5 15.Bxg5 e5 16.Bxf6 Qxf6 17.fxg4 exd4 18.g5 Qf3 19.g6! Bg4!?
White has a clear advantage after 19...fxg6 20.Nxd4! (Lautier-Piket, Cannes 1990). While after 19...Qxh1 20.gxf7+! Kxf7 21.Qf4+ Black, at best, loses a queen for insufficent compensation: 21...Ke7 22.Ng3 Qh2 (in order to have 23...Kd7 after 23.Qg5+) 23.Nf5+ Bxf5 24.Qxh2 Be6.
19...Bg4!? was suggested by myself in my book, and later tried unsuccessfully against Lautier in The Odessa 2006 Pivdenny Bank rapid. The game was a disaster for Black, indeed. After looking at this position again afterwards, I realised that from the beginning I had underestimated the danger of Black's king position. Even with best play White's chances are better.

20.Rg1
Curious but hardly promising an advantage is 20.Nxd4 Qxd1+ 21.Qxd1 Bxd1 22.Bxc4 with some crazy play: 22...Kg7!? and 23.Rg1!? can be answered by 23...Bg4!! (not 23...Rxe4?! 24.h6+! Kf6 25.g7! and White is better!) 24.Nxc6 (Black's basic idea was 24.Rxg4 Ne5!) 24...Rxe4 25.Bd5 Rae8 26.Bxe4 Rxe4 and Black is OK.
The most critical, possibly, is 20.gxf7+! Kxf7 21.Rg1! Ne5! (really a sad necessity... White's very strong threat was 22.Nxd4! Qxd1+ 23.Qxd1 Bxd1 24.Bxc4+, while bad for Black is 21...Bxh5 22.Rg3! with the idea of 22...Qxe4 23.Bg2! Qxe2 24.Rf1+ or 21...c3 22.bxc3 and again there is no solution, for example 22...Ne5 is refuted by 23.Rg3 Qxe4 24.cxd4! or 21...d3? 22.Nc3 threatening 23.Bg2, 24.Rfd1) 22.Qxd4 with an important position, where White is somewhat better. His main idea is Nc3, which can be prepared by Kb1. Possibly, Black's best is 22...Qf6 (there are other options such as 22...Re6 but they fail to impress; note that the following forced line favours White 22...Nd3+?! 23.Kb1 Rxe4 24.Qd5+ Re6 25.Qxc4 Ne5 26.Qxc7+ Re7 27.Qc1! Rc8 28.Nc3 with the idea of 28...Qxd1? 29.Bc4+!) and now White has a choice. 23.Kb1!? (after 23.Nc3!? Bxd1 24.Bxc4+ Nxc4 25.Qxc4+ Qe6 Black possibly holds; still it is hard to be sure... curious is 26.Nd5!? Qxe4 27.Qxc7+ Ke6 28.Rg6+ Kxd5 and somehow Black survives: after 29.Qxd6+ Kc4 30.b3+ he has 30...Bxb3 31.axb3+ Kxb3) 23...Rab8 (activating a rook, at least) 24.Nc3!? (there are calmer possibilities such as 24.Rc1 where Black may try 24...c5!? 25.Qc3 Rb4 and 24.Ka1!? ) 24...Bxd1 25.Bxc4+ Nxc4 26.Qxc4+ Qe6 27.Qxc7+! Qe7 28.Qc4+ Qe6 and again it may seem that Black should hold, but there are no guarantees.

20...Rxe4?
Just a losing move. In some other lines such as 20.Re1 this capture could have been playable, but not here when White is ready for Bg2.
After 20...Ne5!? White maybe has nothing better than 21.gxf7+ . Which is good enough, indeed, and transposes to a line with 20.gxf7+ after 21...Kxf7.
Another possibility is 20...fxg6!? 21.hxg6 (or 21.Re1 c3!? 22.bxc3 and now 22...Bxh5! but not 22...gxh5?! 23.Nxd4 Nxd4 where White successfully attacks by 24.Bc4+! Ne6 25.Rgf1 Qg3 26.Qg5+ Kh8 27.Qh6+ Kg8 and here 28.Re3! Qe5 29.Rf6 Re7 30.Rg3!; in the line 21.Qh6?! Ne5! 22.Rg3 Qf8 Black should be fine) 21...Kg7 (after 21...c3 22.bxc3 Rab8 23.Nxd4! Black has problems) 22.Re1 d3! (probably best, though there are other serious moves) 23.Bg2 (after 23.Rg3 Qxe4 24.Bg2 Qf5 Black is doing well) 23...Qf2! (provoking White's next; less precise is the immediate 23...Qxe2?! ) 24.Rgf1! Qxe2! (favours White 24...Qxg2 25.Rf7+! Kxg6 26.Nf4+ Kxf7 27.Qxg2) 25.Rxe2 dxe2 with reasonable compensation for the queen.

21.gxf7+ Kxf7 22.Bg2 Rxe2
Now everything loses for Black. For example, 22...Qxe2 23.Qxe2 Rxe2 24.Bxc6 and after 24...Rg8 White has 25.Bd5+.

23.Bxf3 Rxd2 24.Bd5+ Ke7 25.Rxd2 c3 26.bxc3 dxc3 27.Rh2 Bf5 28.Bxc6 Rb8 29.Re2+
1–0 Lautier-Golubev, Odessa rapid 2006.

(The game can be viewed at www.chessgames.com/perl/chessgame?gid=1424066).