In this whole game,where did he actually go wrong?

I think 17...f5 was a mistake by black.In an attempt to have a more open game, he simply exposed his kingside to an open attack.

after 29...kg8 30.Nf6+ Nxf6 31 Nh6+ Kh7 32. Nf7#

after 29...kg8 30.Nf6+ Nxf6 31 Nh6+ Kh7 32. Nf7#>

A bit of a calculation mistake on yur side....Black cant play 31...Kh7 as that square is controlled by the bishop on b1.

And if he plays 31..Kh8 32.Nf7 is not checkmate.

I had Fritz 12 running for a little while and it liked black before 19...bxa4. I don't think 19...bxa4 is the sort of move Black wants to play here - it gives white the c4 square?! - and playing down the b-file sounds too slow. White's potential k-side attack can be lethal. I guess 19...bxa4 20. exf5 Rxe3 21. fxe3 Bxd5 22. e4 Bc6 is possible. Fritz calls this equal

Kasparov doesn't mention 19...bxa4 in his notes. Karpov played 19...Qd7 in a later game. 19...c4 has also been suggested. "The Ruy Lopez: a Guide for Black" by Johnson and Johannessen (Gambit 2007) didn't consider 19...bxa4 either.

It seems academic anyway because white has moved away from 18. Rae3. Yes, we are still in opening theory at move 18 in this line. 18. Nh2 Nf6 19. Rf3 has been played a lot.

As for understanding this position? Well, I think these main Ziatsevs are a mess! Perfectly suited to the great Ks but not for us mere mortals. Also, I haven't seen any time when a GM has played bxa4 in these lines.

In video of the match you can see him calculating, on this as on later moves. He'd seen a stupendous, dazzling array of tactics -- not a single forcing line, but a whole range of different ones, starting with the truly beautiful stuff if Black takes the Knight on its first visit to h6: 26.Nxh6! Rxh6 27.Nxd6 ... and he had winning attacks prepared for 27...Qd7, 27...Qh5, even 27...Qxe1+.

Okay, so he hadn't seen *everything* through to a win, and there was intuition involved in playing 26.Nxh6. But he didn't just blitz it out because it looked nice. He'd seen deeply into the position.

His slight slip-up near the end, playing the winning Bxg6+ rather than the crushing Rxg6, is trivial, and to criticize Kasparov for 'missing a mate' is just sad. Yes, Rxg6 is a pretty win and might have given the game the classic status it deserves. Mate in 3 or 4, depending which desperado lines Black plays - one nice sequence is 37.Rxg6 Ne7 38.Rxe7 b1(Q) 39.Ng5+ Kh8 40.Rh7#. Of course 39.Bxb1 also mates, but letting Black promote is more pleasing.

Of course 37.Bxg6+ is also a pretty trivial win. He'd done the heavy lifting by then. I remember once being congratulated by an IM for winning a game with a sacrifice. "Oh", I moaned, "I should have won quicker". "Yes", he replied, "You had a mate in four and you played a mate in six. So what?"

This K-vs-K epic was a key game in arguably the last great title match in history: later events have been one-sided or too short. The reason Kasparov's opening prep became legendary is that he honed it in long matches against a very dangerous rival -- after 20+ games, a player might have to use their third-choice defence if the others were looking suspect. It doesn't happen in 12-game matches.

At several points he flashed out a quick sequence of moves, said "This wins" and moved on. Each time I'd reconstruct the sequence, find the 'final' position, and have to *work* for several minutes to see a win. With the final position on a board in front of me -- a position Kasparov had visualized and evaluated six or eight moves earlier...

<11.g5>

click for larger view

is an often played move that can lead to a quick draw. Black has no better answer than <11...f8>, then White has nothing better than moving back with <12.f3>. There are literally dozens of games in the database where the players just repeat these moves and quickly agree to a draw.

The idea can also be used as a device to play a few extra moves without thinking in order to save time on the clock.