Even GMs Stumble on K+P Endgames

Continuing with Smyslov's Sparklers, the first diagrammed position occurred almost 20 moves after the position discussed in Which Side Is Better. Kasparov started his notes on the game here.

Euwe played 33...Bd2 and eventually lost. Smyslov gave a long analysis of the alternative 33...Bg5, trading Black's bad Bishop for a Knight. The object of the analysis was to determine if the move is sufficient to draw. After 33...Bg5 34.Kb3 Bxh4 35.gxh4 Ne7 36.Nxe7 Kxe7 37.Kc4, he reached the position shown in the second diagram.

Groningen 1946
W: Smyslov, Vasily
B: Euwe, Max
[FEN "6n1/2p2k2/3p1p1b/3PpN1p/4P1pN/2P3P1/1K3PP1/8 b - - 0 33"]
(After 33.Kc2-b2(xP))

(After 37.Kb3-c4 (analysis))

King and Pawn endgames are notoriously tricky, and the diagrammed position is no exception. Smyslov gave 37...f5! (37...Kd7 loses) 38.exf5 Kf6 39.Kb5 Kxf5 40.Kc6, and established that both 40...Ke4 and 40...g3 are insufficient to draw. He didn't discuss 40...Kf4, which looks better than the alternatives.

Kasparov copied Smyslov's analysis, but gave a different set of variations. It appears that Smyslov changed his published analysis of the position, but the English translation of his book used the original notes.

After 37.Kc4, Kasparov gave 37...f5 38.f3 (Not 38.exf5 Kf6 39.Kb5 Kxf5 40.Kc6 g3! 41.f3 e4! 42.fxe4+ Kxe4 43.c4 Ke3 44.Kxc7 Kf2 45.c5 Kxg2 46.cxd6 Kf3 47.d7 g2 48.d8=Q g1=Q. [MW: After 47...g2, Smyslov terminated his analysis with 'etc.'; Kasparov didn't say why the position is not lost for Black after 48...g1=Q. In any case, 40...Kf4 needs attention.]) 38...f4 39.Kb5 Kd7 40.c4 Kd8 41.Kc6 Kc8.

Now another curiosity appears. Kasparov continued 42.c5 dxc5 43.Kxc5 Kd7, when White wins after a sophisticated maneuver that looks like triangulation. In fact, the game is won easily after 42.fxg4 43.h5, when White promotes first and checkmates Black immediately.