It's not perfect though. Kramnik I believe lost thread of the game early on during the sequence

18. Bg5 Nc5 19. Ba2 h6 20. Bh4 g5

Some comments:

19. Ba2?

I don't quite understand why Kramnik would not want to exchange off the strong Black WSB on e4 by Bc2.

20. Bh4?

I think this is a clear positional mistake. After g5, the White DSB has been practically locked out of the game.

Now all that Hare has to do is to remove his Knight from f6, and push his f-pawn f5 and f4, and White's bishop would have been transformed into a big pawn. Black would be winning positionally.

I thought at this point that Black has achieved an almost decisive advantage, thanks mainly to Kramnik's strangely poor play.

But then Kramnik turns it around. He must have seen that he has drifted into a practically lost position, and came up with the brilliant and fantastically imaginative sacrificial maneuver that everyone is talking about.

23. Rad1 Nfe4 24. Rd5 f5 25. Rxe5 dxe5 26. Bxe5+

I think White is still objectively lost, but this rook sac was Kram's best practical option, and he courageously took it! White's dark squared bishop that was about to be trapped suddenly turns into a monster, supplementing his WSB in raking the open diagonals across the center. Kram also manages to grab two and later three pawns for his rook.

As a human being, I would actually prefer playing White at this point. It's easier to find reasonable moves, and there is a ready-made plan that plays itself: just push up your three connected Queenside pawns if Black dilly-dallies around. Even though a computer would surely beat White's position, we must remember that it was two humans that were playing the game. Hare actually played well for many moves, in spite of what must have been the tremendous pressure he was feeling.

Remarkable concept. Kramnik judged that the continuous pressure of his queen and bishop pair against Blacks weakened kingside plus his pawn trio on the queenside was more than enough compensation for a whole rook.

<HeMateMe: greatest "proponent"?>

He sure has the longevity though. There are very few "old folks" who used to be in the top echelon when they were younger and are still up there, kicking it with the young uns.

5. d3 is the Ruy Lopez Anderssen with a quiet center; this game went 5. 0-0 Be7 6. d3 which is similar. After 9. Nc3 it is more of a Spanish Four Knights situation as usually White's QN goes to d2. Since White won, pundits will claim that White's Rook sacrifice was bold while Black's Kingside Pawn play was reckless; had Black won, expect the reverse. Note White has lots to do through advancing the Queenside Pawns, meanwhile Black had little to do.

Some insightful commentary:

<25. Rxe5!!?? After spending 11 minutes, a gutsy call by Kramnik. Just awesome assessment of the resulting position, even though I feel all lines could not have been evaluated. Hari must have felt a lot of pressure, using 14 min for 26 ... Nf6 and several more minutes on other moves in this sequence.>

<The first half of this game looks like a typical Spanish with black defending and reaching what looks like an equal game. Suddenly white sacrifices a rook for open lines to black's king and three passed pawns, which however are not very far advanced. After this positional sacrifice both sides play logically but black (in time trouble probably) misses 33...Qa6 attacking a3 and c4, white's pawns move forward and in a few more moves black is lost. This is a creative win for white, even if analysts find a refutation this would be extremely hard to do at the board.>

<Fantastic rook sac by Kramnik. Certainly Harikrishna made some subpar moves, like 33...Qc6 (33...Qa6 is better), 34...Rbd8 (34...Kh7 is better), and finally the fatal 36...Kf8? (36...Qb3 was essential), but I am not clear that Black ever had a winning advantage, in spite of the extra rook.>

<I think 33...Qa6 is not easy to play, as it isn't so simple as attacking c4 and a3: the point is presumably to take on c4 with the bishop and avoid getting pinned by Rc1, as would have happened with the queen on c6 (taking on c4 with the queen is bad because of Rc1-c7). Sounds nice, but this is a passive placement of pieces, and 33...Qa6 34.Qb2 Bxc4 35.Ne5 is unsettling to look at for Black, even if the computer decides there is nothing for White.>

<I think White is still objectively lost, but this rook sac was Kram's best practical option, and he courageously took it! White's dark squared bishop that was about to be trapped suddenly turns into a monster, supplementing his LSB in raking the open diagonals across the center. Kram also manages to grab two and later three pawns for his rook.>

<As a human being, I would actually prefer playing White at this point. It's easier to find reasonable moves, and there is a ready-made plan that plays itself: just push up your three connected Queenside pawns if Black dilly-dallies around. Even though a computer would surely beat White's position, we must remember that it was two humans that were playing the game. Hari actually played well for many moves, in spite of what must have been the tremendous pressure he was feeling.>

<Remarkable concept. Kramnik judged that the continuous pressure of his queen and bishop pair against Blacks weakened kingside plus his pawn trio on the queenside was more than enough compensation for a whole rook. It must be noted also that after some 15 years without playing 1-e4 , Kramnik has been achieving great results from that!>

Csom vs Yusupov, 1982

41.? White to play and win