(Well, I often thought that a few more seconds of careful reading would have helped a lot in my live - Lesen müsste man können).

This article makes the statement, "And then, instead of e.g. 23.cxd6, he uncorked 23.Nfxg5!!?. At first the move simply looked bad, by which I mean the computers were horrified, but then it gradually emerged that White does have enough compensation for the piece – though no more. 30.h4 was perhaps a better means of forcing a draw."

The reference to computers already reeks of unprofessional arrogance, making the statement just plain wrong. The position is too complicated, engines would show much here (certainly not horror) under short term computation. Indeed, here is the output of my computer running Houdini_15a_x64 with multiple threads (nearly 4 million evals/sec):

21/56 03:22 768,937,104 +0.52 23. ... hxg5 24.Qxg5
22/60 06:04 1,417,737,752 +0.53 23. ... hxg5 24.cxb6

It computes the move as giving a slight advantage to black, most probably still a draw. I was not present during this move, I got to the game on move 31 for white. Looking at earlier kibitzing on here, apparently the move produced quite a bit of discussion amongst the kibitzers even GM Ray Keene decided to comment with <ray keene: i never use a chess computer while commenting-i prefer to use my own brain and let others tell me what the machines think> and I wholeheartedly agree.

I was first to kibitz on 32.Ng5 first thinking it dubious at best but after more thought seeing that it probably loses. The article states, "The magnetic lure of the g5-square again drew his knight: 32.Ng5??! with 34.Ne6+ following. Although playing almost a full rook down Nakamura did leave Kramnik a last chance with 35...Bf6: (diagram). Here 36.Qg6!! instead of Kramnik's 36.g4? might still have made the game a minor masterpiece."

I had commented 35...Qf6 was best and then commented 35...Bf6 was weak. 35...Qf6 gives more luft to black's king, double protects the B on g7, attacks the pawn on f3 and (presuming 36.Ng5 to attack the rook on f7) allows 36...e4! to reinforce black's attack.

However, 35...Bf6 although much a lesser move, is far "making the game a minor masterpiece" black still has some slight advantage after 36.Qg6 with 36...Qd7 37.f4 Ke7 unpinning the rook for a likely draw.

At the juncture, both players moved quickly. It was not easy to see how weak 35...Bf6 was or how poor 36.g4? was. The only thing obvious to me was that white's plan to run his pawns was way too slow, and black had winning counterplay via 36...e4 to first exchange pawns and place the bishop beautifully on e5, but then black's choice of 36...Qb7 to swing to queen around was also clearly very good (winning).

Looking back with more time and Houdini to verify, 32.Ng5? was clearly a losing move but afterwards both players made mistakes with 35.Bf6?! 36.g4? with white still losing probably faster. White's sac on move 32 was unsound either way.

White may have a winning positional advantage in the position before 23 Nf3xg5, so that this move can be said to start a sound sacrificial attack, an attack justified by sufficient positional advantage.

32 Nxd6 instead of 31 Ng5 might have drawn but could Kramnik have won?

I would like to point out that it was said of Kasparov that he brought his last piece into play before beginning a final attack.

Houdini does indeed evaluate 31.Raa7 as best but drawn because of the exchange 31...Rxe7 32.Rxe7 b5 33.g4 Qf8 34.Qxg6 Qxe7 35.Ng5 and the perpetual via Qa7-Qa2 from there.

This was similar to the perpetual you and others were referring to after 31.Qxg6 with 31...Rxe7 when I started getting into the position and suggested what I did then (31...Qxe7 what Naka played).

<The reference to computers already reeks of unprofessional arrogance, making the statement just plain wrong. The position is too complicated, engines would show much here (certainly not horror) under short term computation.>

I assure you I have absolutely no arrogance whatsoever when it comes to chess - more like an entirely justified inferiority complex :) And I think it's pretty clear to anyone that a statement like "the computers were horrified" isn't intended to be taken entirely seriously. Computers don't feel any emotions whatsoever. Nevertheless, all our computer friends (also not serious, I'm not friendly with any computers) switched from giving White a tiny edge to clearly favouring Black. e.g. Chessbomb now has -0.84, Whychess is -0.91, Chessok is now only -0.22, but all of those figures were initially much higher in Black's favour. So as a description of the initial "computer" reaction I think what I wrote's fine, just intentionally hyperbolic. Of course I don't think humans can outplay computers in these positions, but still, Kramnik's calculation or intuition was actually better than their first judgements - which is something.

On the "minor masterpiece". Maybe I should have been clearer, but I don't mean by that White would win - getting a draw after playing those two sacs/dubious lines would strike me as a minor masterpiece (far from flawless, but what is?). Overall, as far as I can see, you saw the game pretty much as I did.

<The article's attempt to try and somehow relate these earlier games to this one is some sort of dramatic rubbish.>

No, I do not think you were being arrogant, forgive me for that. I guess I was still annoyed with some comments by others, resulting in me giving 3 IGNORES and getting distracted, from their rude accusations about engines during the game.

Thanx for compiling the article anyways, as I wrote to begin with it was (otherwise) very informative. Yes, you seemed unclear about what "minor masterpiece" meant, it certainly implied to me white was going to win with the sac (or that it was even good).

In all honesty, I claim less than average ability with computer chess, it is only recently I started using engines as much other than an opponent. I just know it takes them awhile to evaluate complex positions such as these and do not trust them during live games except as fun inputs (like GM Keene posted).

It's lousy see engines are abused or a few making rude comments to good players accusing them of abusing engines. Personally, not knowing others on here, I much prefer to assume they understand chess quite well in making their posts unless they convince me otherwise (such as "Life Master" AJ with his usual arrogant rubbish in Kasparov vs Topalov, 1999 few pages up).

I *could*, but I won't. Not because they don't exist, but because you've already been told by others where to find them if you're actually interested, and not just doing your gadfly routine.

I think this and his other comments show he did it for the reasons others and myself listed, and it was a calculated descision. He even said he didn't want another draw, and if he played as in times past he doubtfully would have left with such a strong result it's easy to judge from after the game but how many of the same people would be on him for not trying to play risky to defeat a weak Nakamura with white?

click for larger view

I see only two ways Black can defend against the threat of 37. Qg8+:

<A> 36. ... Qb7 37. Ra2 (the purpose of this move is to avoid the gaze of the bishop on f6, for example: 37. h4? e4! 38. Re1 Qxd5 39. Qg8+ Ke7 40. Qxc8 Qxe6 41. Rxe4 Be5 ) 37. ... Ke7 (37. ... Qxd5?? 38. Qg8+ and it is mate in 17 according to my engine!!) 38. h4

1) 38. ... Ra8 39. Rc2 Rc8 (My engine's suggestion might actually be losing for Black: 39. ... Ra7?! 40. Rc7+ Qxc7 41. Nxc7 Rxc7 42. h5! and I think the h-pawn and the immobility of Black rooks gives an advantage to White.) 40. Ra2 =

2) 38. ... b5 39. Kg2 Qb8 (Black can't play the planned 39. ... Qb6 b/c of 40. Ng5. Black is worse/losing in every line: 40. ... Rg7 41. Qf5! note that the c8-rook is undefended; 40. ... Bxg5 41. Qe6+ Kf8 42. Qxc8+ Bd8 43. Ra8; 40. ... Rcf8 the safest option 41. Nxf7 Rxf7 42. h5!) 40. h5 (and here in almost every single line my engine gives a perpetual)

It seems that Bronstein basically had this exact feeling in his games all too often, with mixed results...