This game won the best game award in Informator 67 with 87/90 points. The second best game (also a Kramnik game) received 55/90 points.

btw, at no point was Kasparov winning. 23. Nf3? instead 23. Ra2 Be4 24. Re1 Bb7! 25. Qd3! ; and 24.Nc5? instead 24. Qe2! holds the balance.

Conrad93: To be fair, Kasparov played terribly right from the opening...>

It does not matter if the engine does not like Kramnik's position at move 22, as over the board it is hard to refute it unless you are skilled defensively.

Kramnik is no less talented than Kasparov positionally and strategically. Kasparov excels in creating and calculating deeply in sharper positions where he has the initiative, much of which is aided by his opening preparation.

click for larger view

BLACK TO MOVE
In his annotations (Informant No.66) Kramnik attached an exclamation mark to the move 22...Qh3!, and added lengthy analysis intended to prove that 22...Rxf1+?! would have been a dubious move leading to White's advantage. What does Stockfish 12 have to say about that? Well the opinion of the engine is that White stands better in the diagrammed position, and it recommends 22...Qe5 as the best move for Black. LcZero shares the same opinion. In fact, the engines don't seem to like Kramnik's move 22...Qh3?, on account of the reply 23.Qe2! with a high evaluation in White's favor, namely +2.86. LcZero shares this point of view. A sample variation: (from the diagram): 22...Qh3? 23.Qe2! Rxf1+ 23.Qxf1 Qxf1+ 24.Kxf1 Bxe4 25.Bf4 Rf8 26.Ne6 Rf6 27.Nac5 Bxc5 28. Nxc5 Bd5 29. Ke2, and White has a healthy extra piece and a won ending (computer evaluation: +3.63) (see diagram below):

click for larger view

click for larger view

WHITE TO MOVE
Kasparov played 24.Nc5?, and Kramnik attached a question mark to this move (Informant 66). Moreover, he attempted to show that 24.Qe2! was the correct move, and added a tremendously lengthy analysis with tons of variations and sub-variatons. (I wonder if ever there was a person on our planet who played over those lengthy variations with a chessboard and pieces). But is 24.Nc5 really a bad move as Kramnik assures? Curiously enough, the engines do like the move 24.Nc5! and their veredict is that the position after that move is about equal (evaluation: -0.30). As a matter of fact, White's decisive mistake came just a move later: 24.Nc5! Rxf3 25.Rxf3?? That move was a blunder, changing a computer evaluation of 0.00 to one of -4.21. It is curious that Kramnik did not attach even one question mark to the move 25.Rxf3. He only mentioned that a better alternative would have been 25.Ra2, and added lengthy analyisis intended to prove that Black would have had a better position anyway. The engine, however, has a different opinion. A sample variation (from the diagram): 24.Nc5 Rxf3 25.Ra2! Rxf1+ 26.Qxf1 Qxf1+ 27.Kxf1 Rc8 28.Be3 Bf4 29.Bxf4! Rxc5 30.Ra4 (diagram), and according to all engines the position is equal and a draw as a likely result notwithstanding Black's extra pawn.

click for larger view

Kramnik has a different opinion, and wrote in his notes that Black has a clear advantage in the diagrammed position. Nobody should wish to argue with the former world champion, so let us take it for granted that Black has a clear advantage in the diagrammed position; but just for the sake of having fun, let us have this ending played out by SF12 vs. SF12 at 10 seconds per move: 30...Bxe4 31.Rxb4 Bd3+ 32.Kf2 Rc2+ 33.Kg3 Bb5 34.h4 Re2 35.Rd4 Re1 36.b4 Kf7 37.Bd6 (evaluation: -0.07) h6 38.Rf4+ Kg6 39.Bc5 Bd7 40.Rd4 Bb5 41.Rg4+ Kh7 42.Bd4 Re7 43.Kf2 Bc6 44.Rf4 Be8 45.Bc5 Re5 46.Bd4 Re7 47.Bc5 Re6 48.Rg4, draw by repetition; computer evaluation: 0.00. (DIAGRAM)

click for larger view