After 32 Ne5 Black's f7 pawn is attacked but it also has the task of defending the g6 pawn so that on 32...f5 the g6 pawn is attacked by the N on e5.

After 33 Nf6 Kramnik moves the g6 pawn by 33...g5 but the tempo spent on that move gives Karjakin's N on e5 time to start an endgame attack by 34 Nc6+, an attack which ends up winning a pawn on g5.

However if Kramnik can draw at least by 39...fxg5 as <Polarmis> has indicated this suggests that White's endgame attack was either mistaken or mistakenly executed.

houdini says kramnik is probably okay around move 42 but then starts to implode!Sergey Karjakin - V Kramnik, Russian Superfinals 2011

click for larger view

Analysis by Houdini 1.5a w32:

1. (0.70): 42...Bd7 43.Ne5 Bc8 44.Ke3 Ne8 45.Ngf7 fxg4 46.hxg4 Nf6 4

2. (1.26): 42...fxg4 43.hxg4 Ba6 44.Ne5 bxa4 45.bxa4 Bc8 46.Ne4 Ne6+ 47.Ke3 Kb6 48.g5 Ng7 49.Nd3 Ba6

Not sure why so many insist on thinking Karjakin had some big brilliance here or that Kramnik did not know what he was doing with 39...c3 either. I mean, we are talking top GMs (one a recent WC!) and, as shown above, 39...c3 not only draws but it was probably also black's best chance to try for a win (at greater risk than simply 39...fxg4 the obvious drawing move).

Kramnik had the right idea but he got too fancy with his execution of it, and messed up, simple as that. Hence his reported look of disgust and quick exit when young Karjakin started going over 39...fxg4 with him, as if he did not know this.

While Shipov's "big idea" of exchanging rooks as "being bad" does lead to a very slight advantage for white in a computer eval, a few moves later white chose 32.Ne5 instead of 32.Nf6 as well, kicking that eval back to dead zero. Neither thing "was bad" here, Kramnik exchanged rooks with the idea of possibly taking best advantage of his extra pawn later on in the endgame. And Karjakin simply wanted to better activate his knight.

I liked this interchange between Grischuk's suggestion and Karjakin's willingness to try to mate with two knights, as that answered my main question whether the game was lost on move 39.

<Grischuk: “What about 41...fxg4 42.hxg4 Nf5+ here?” Then 43.gxf5 Bxf5 44.Nf3 Bc2 45.Nd2 b5 46.Ne5 bxa4 47.bxa4 Bxa4 48.Nb1 Bb5 49.Na3 – objectively that’s won for White. I was ready to play that position with “two knights versus pawns”, as I had no choice. It seems that was Black’s last chance.>

"39...c6 Correct was 39...fxg4 40.hxg4 Nf5+ 41.gxf5 Bxf5 42.Nxg5 Bc2 – I’m not losing, but I’m not winning either – a draw. It’s very strange that he missed that chance as he thought for 10 minutes here. Probably he simply didn’t see the idea." -Karjakin

From the added info that he took 10 minutes on it, it's clear he thought through and decided on 39...c6 as better, giving more chances. But as I stated, he got too fancy, postponing 40...fxg4 too long. This position is not complicated.

Karjakin vs Kramnik, 2011 40 ... Ng7-f5+! intends 41 g4xNf5 Bc8-f5 then ... Bf5-c2-xb3xa4

Kasparov vs Kramnik, 2001

In both games, white makes use of his two main opening trumps as follows: sacrificing the e-pawn to expose the awkwardness in black's position. Winning a pawn back on the kingside. Then finishing with a pure Ruy Lopez Exchange strategy where white's kingside majority can pass its own pawn, while black's disrupted queenside majority cannot.