This is the identifying move of this gambit, which was introduced at about the same time by Rasa in New Zealand and Studier in Germany.4...Bb4 5.f3 c5 6.e5 Bxc3+ 7.bxc3 Nfd7 8.f4 Nc6 9.Nf3 Qa5 10.Qd2 0-0 11.Bd3 c4 12.Be2 Nb6 13.0-0 Na4
Black plays to win a pawn while his Kingside is lightly defended.14.Qe1 Nxc3 15.Bd2 Nxe2+ 16.Qxe2 Qd8 17.c3 Ne7 18.Ng5 h6 19.Qh5!? f5!
On 19...hxg5? 20.fxg5 Black would have to give up material to prevent mate along the h-file.20.Nf3 Bd7 21.Kh1 b5 22.Rg1 Be8 23.Qh3 Bg6 24.g4 Qe8 25.g5 h5 26.Bc1 Nc8 27.Nh4 Kh7 28.Ba3 Rf7?
Here Rasa commented: "Black has a poor choice between Rg8 and Rf7. The text turns out to be the beginning of the end, but who could have foreseen that at this stage?"29.Nxg6 Kxg6 30.Rg3 a5?
Rasa wrote that Studier was under the impression that 32...Ne7 followed by Nxg6 would force a draw, but this was refuted by a New Zealander, W. Petrie, who gave 32...Ne7 33.Rag1 Nxg6 (33...Kh4 , however, still gives Black a chance to draw: 34.gxf7 Qxf7--tvp) 34.Rh3+ Nh4 35.Rg5+ and mate on the next move.33.Rag1 Kh4 34.Rf3 Ne7 35.Rgg3 1-0.
35.Rgg3 Nxg6 36.Rh3+ Kg4 37.Rfg3+ (37.Bc1 also mates.) 37...Kxf4 38.Bc1+ Ke4 39.Re3+ Kf4 40.Ref3+ Ke4 41.Rf4+! Nxf4 42.Re3#Update: Of course all of the above analysis was done without the aid of computers. It might be interesting to take the time to check it thoroughly with one today. I believe Black could survive after 19...hxg5, and also that White should get at most a draw after the Queen sac. "Twenty years ago we were doing things that don't work today because of computers. We used to bluff our way through games, but today our opponents analyse them with a computer and recognize in a split second what we were up to. Computers do not fall for tricks." GM Anand in a recent Spiegel interview, translated at ChessBase.com