"Garry's Choice": A Great Predecessor
In the preceding post I reviewed Informant 113 and mentioned a new column, "Garry's Choice". In the column's initial installment Kasparov features the game Paragua - Debashis, taking special note of a spectacular missed possibility:
Black has enough extra material to win several games, but his king is in a world of trouble. In the game he played 24...Kf8, which allowed a forced mate, but what he missed was 24...Qg4!!, after which the best White can do is play 25.Rxg4+ Kxf7 26.Qxh7+ Ke6 27.Re4+ Kd7 28.Rxe7+ Bxe7 29.h4, with good drawing chances.
An even prettier version of that move was possible a couple of moves before.
Here White played 23.Bxf7+, but 23.Rg1 is more accurate and more frightening, cutting off the enemy king's escape.
The only move, as you've surely surmised by now, is 23...Qg4!! The queen can be taken four different ways, even with check, but in every case Black is at least equal.
Kasparov confesses that this move is unique to him, and the best he can do to come up with a vague predecessor is "Mitrofanov's Deflection", the crowning blow to a deservedly famous composition. The key moment comes here:
White plays the spectacular 1.Qg5!!, pulling Black's queen to a dark square, so that after 1...Qxg5+ 2.Ka6 it can't safely check White's king along the f1-a6 diagonal.
The other predecessor Kasparov suggests is even less compelling, so while I'll provide it in the replayable games section I won't bother with it here.
Nevertheless, while the (missed) ...Qg4 idea is indeed magnificent, it's not entirely unique, and Kasparov missed a far greater predecessor. First of all, it's from an actual game. Second, it wasn't found later on and possibly by a computer; it was found by the player himself and executed in the game. Third, it's far more similar, making it a genuine predecessor.
MacDonald-Burn, Casual Game 1910, position after 33.Bh5
White is down a piece for a pawn, but the bishop on g5 is a goner and Black's king is looking kind of crispy. But once again her majesty comes to the rescue: 33...Qg4!! and all is well (or at least almost all). The same piece, going to the same square, and for at least one of the same reasons - to obstruct the g-file. Another similarity is that all White captures but one give Black an immediately winning advantage.
MacDonald correctly played 34.Rxg4, and after 34...Nf3+ the accurate 35.Kg3 would have maintained some advantage. After 35.Kg2? Nxd2 36.Rxg5+ Kh6 Black was better (in the 35.Kg3 case White would have 37.Kg4) and went on to win.