in-depth analysis of game three

Nov. 16, 2003 – There isn't a whole lot to say about the way the game progressed. X3D Fritz never found a plan and Kasparov did whatever he wanted on the queenside. The story, then, is why X3D Fritz was unable to find a plan and what Kasparov did to keep the mighty machine in the dark.

[Replay the annotated game online and download PGN ]

Kasparov,G (2830) - X3D Fritz [D45]
Man-Machine World Championship
New York City USA (Game 3), 16.11.2003

1.Nf3 Nf6 2.c4 e6 3.Nc3 d5 4.d4 c6 5.e3 a6 (D1)

Diverging from game one, when the more common 5...Nbd7 was played. This sideline of the Slav with 5...a6 was criticized by Kasparov after the game. From the continuation here we can see why.

X3D Fritz is playing from its opening book right into a position it doesn't understand at all! When Kasparov said this everyone looked over at poor Alex Kure, the man responsible for selecting and "training" X3D Fritz's openings.

Kasparov has little practical experience with this sideline and in both cases (once with white and once with black) the game continued with 6.b3.

When this move appeared on the screen I thought maybe the Fritz team were trying to give Kasparov a little psychological jab. He played this move himself a month ago against Huzman in a game that turned into the shortest loss of his career after a horrific blunder. They wouldn't admit it, but you know they knew about that game and that Kasparov would be forced to think about while he was sitting there against X3D Fritz.

6.c5 Nbd7 7.b4 a5 8.b5 e5 [8...Ne4? 9.Nxe4 dxe4 10.Nd2 f5 11.f3 Qh4+? (11...exf3 12.Qxf3+/-) 12.g3 Qh6 13.Qe2 1-0 Euwe,M-Alekhine,A/NLD 1935/(41); >=8...e5]

9.Qa4! Given an exclamation point by Gligoric and Wade in their book "The World Chess Championship" (1972). This move was apparently not in X3D Fritz's opening book so it was now on its own.

9...Qc7 10.Ba3 [10.Be2 e4 11.Nd2 g6 12.Nb3 Bh6 13.Bd2 0-0 14.0-0-0 b6 15.bxc6 Nb8 16.cxb6 Qxb6 17.Qb5 Qxc6 18.Qxc6 Nxc6 19.Na4 1-0 Pachman,L-Fichtl,J/Prague 1954/MCD (41)]

10...e4 11.Nd2 Be7 All of these moves had been played before, although Kasparov said afterwards that he was not conscious of that at the time at this point. The game they are following was not a minor one, but a battle from the world championship match-tournament in 1948 between American Samuel Reshevsky and Soviet (Estonian) Paul Keres.

12.b6 Immediately sealing the queenside and diverging from that game from 1948, although there are still many similarities. A player as strong as Keres had no difficulty in seeing that Black needs immediate pawn play on the kingside in order to compensate for White's advantage on the kingside. He played a rapid ...h5 push and entered a very sharp battle.

[12.Be2 h5 13.b6 Qd8 14.h3 Nf8 15.0-0-0 Ne6 16.Ndxe4 Nxe4 (16...dxe4 17.d5 Bxc5 (17...0-0 18.dxe6 Qe8) 18.dxc6 Bxa3+ 19.Kc2) 17.Nxe4 h4 0-1 Reshevsky,S-Keres,P/NLD/URS 1948/MainBase (63) (17...dxe4 18.d5) ]

12...Qd8 13.h3 An odd little prophylactic move that shows that Kasparov is betting that X3D Fritz won't know what to do in this position without any clear targets for its pieces.

All the time Kasparov spent training with X3D Fritz clearly paid off. 13.h3 doesn't develop anything but it takes away a square from Black's knight, the g4 square.

13...0-0 14.Nb3 (D2)

Kasparov closes in on the isolated a5 pawn. After he captures it the white pieces will be a little tied up and during that time Black needs to counterattack vigorously on the kingside.

Instead, from now on we watch Kasparov consolidate on the queenside while X3D Fritz does absolutely nothing on the kingside. It has no clue that its only hope is to play its kingside pawns forward to break through the white pawn chain at its base.

14...Bd6?! This got a good laugh from the Grandmaster commentators and the audience. Only a computer! It puts its bishop right where the white pawn can capture it. If Kasparov takes the bishop he loses his queen after 15.cxd6?? Nxb6 and the white queen is trapped.

Of course Kasparov isn't going to blunder his queen away, so did this curious move have any other value? Maybe so, if Black thinks its bishop is more useful on the b8-h2 diagonal, attacking the kingside.

[14...Ne8 15.Rb1 f5 Here is the key move that X3D Fritz never wanted to play. All of Black's hopes are pinned on eventually breaking through with ..f5. 16.g3 g5 Necessary to enforce ...f4, but X3D Fritz has been taught not to move the pawns in front of its king. Now a double-edge battle is underway and White will have to watch out for Black's breakthrough on the kingside. In the game, Kasparov never had to worry about this at all since X3D Fritz never touched its f-pawn.]

15.Rb1 Kasparov ignores X3D Fritz's provocative play and continues to develop his pieces. Black isn't threatening anything. [15.cxd6?? Nxb6; 15.Nxa5 Nxb6 16.cxb6 Bxa3 17.Qxa3 Qxb6]

15...Be7?! Oh boy, now you know we're in the land of computer chess. As one of America's top players, GM Gregory Kaidanov, put it after the game, "this move showed that the computer doesn't feel any embarrassment!" X3D Fritz puts its bishop right back where it was two moves ago, basically making Kasparov a gift of two moves.

16.Nxa5 Nb8 17.Bb4 Kasparov will slowly unravel his pieces on the queenside and prepare to push his a2 pawn up the board where it will break through and give him a protected passed b-pawn with an easily winning position. X3D Fritz can't see this coming at all and does nothing but watch.

17...Qd7 18.Rb2 (D3)

This useless-looking move confused most of the commentators, but to anyone with extensive anti-computer chess experience it makes perfect sense. The rook protects the f2 pawn, a potential weak spot, but why would you protect something that isn't being attacked?

The reason goes into how computers think. Its brute force calculation can only go so deep, even with four super-fast processors. Black's only possible source of counterplay in this position is to push its f-pawn and open up an attack against the area around the white king, f2 in particular.

If X3D Fritz's search, usually running 12-20 half-moves deep, ever reaches a position in which it sees success in such an attack it will put such a plan in motion. On the other hand, if it cannot reach a favorable position in its searches it will never play the initial moves required. With the rook on b2 protecting f2 already, the potential weakness of that critical square is somewhat hidden from the computer's search.

X3D Fritz can't just play it anyway like a human would, knowing that everything else is useless. A machine has to receive a positive evaluation from its search to play a move and always plays the move that gives it the best evaluation. Since X3D Fritz sees no danger here for itself it is content to play moves that do nothing, but don't cause any negative effect either. It twiddles its virtual reality thumbs. Any human would say, "I have to do SOMETHING."

18...Qe6 19.Qd1 Getting the queen out from behind the pin on the a5 knight. All of Kasparov's moves are based on supporting the push of the extra a2 pawn, with the occasional need to protect against an X3D Fritz threat.

19...Nfd7 After this there was a brief hope that X3D Fritz had found the need to play its f-pawn at long last. 20.a3 Qh6 21.Nb3 Bh4 A pathetic one-move threat that ends up wasting more time. This move pins the f2 pawn against the king and so threatens ...Qxe3+ on the next move.

22.Qd2 Protecting against that threat and preparing to evacuate the king to the queenside. 22...Nf6 Nope, no f-pawn push. Black is doomed. 23.Kd1 Be6 24.Kc1 Kasparov has all the time in the world. 24...Rd8 Useless.

25.Rc2 Nbd7 26.Kb2 Nf8 Ironically, X3D Fritz was reaching incredible search depths because there are so few legal moves in this closed position. It was like casting a powerful searchlight into a black hole. Even reaching 19 half-moves ahead it couldn't find the essential plan.

27.a4 The a-pawn begins its march. 27...Ng6 28.a5 Ne7 (D4)

After all this silliness it's too late for X3D Fritz to do anything now even if it realized it was in trouble. Its pieces are all on the other side of the board while Kasparov crashes through with his pawn.

29.a6 Kasparov gives back the pawn temporarily in order to gain a protected passed b-pawn and squares for his pieces. He will now build up his forces for the final assault.

29...bxa6 30.Na5 Rdb8 31.g3 White gets ready to get his last piece into action and further restrain the useless black pieces.

31...Bg5 32.Bg2 Getting out of the way of the rook while threatening to win a piece with h4, trapping the bishop. [32.h4?! Ng4 33.Bg2 Bf6]

32...Qg6 33.Ka1 Kh8 Two useless-looking king moves that aren't the same at all. Kasparov is getting out of the way of his heavy pieces. X3D Fritz is simply wasting more time. At this point the X3D Fritz team members started to shuffle their feet nervously. They knew from looking at the evaluation that the program had no idea it was about to be crushed.

34.Na2 Heading to the b4 square. 34...Bd7 35.Bc3 Ne8 36.Nb4 Kg8 37.Rb1 Bc8 38.Ra2 Bh6 39.Bf1 Kasparov has optimized his forces for the final strike. During the last 20 moves X3D Fritz has accomplished absolutely nothing.

39...Qe6 40.Qd1 With this move Kasparov's last worry disappeared. He had reached the time control on move 40, which meant he had an extra hour added to his clock. They would get even more time at move 60, but nobody believed the game would reach that point!

40...Nf6 There is nothing to be done at this point, although X3D Fritz was still giving White just a tiny plus. 41.Qa4 (D5)

41...Bb7 [41...Kh8 Waiting passively doesn't work either. 42.Nbxc6 Bd7 This pin backfires. 43.b7! Ra7 (43...Nxc6 44.bxa8Q Rxa8 45.Rb6 Nxa5 46.Rxe6 Bxa4 47.Rxa6 Rxa6 48.Bxa6 Nb3+ 49.Kb2) 44.Nxb8 Bxa4 45.Rxa4 White wins easily even without his queen thanks to the mighty pawns.]

42.Nxb7 Rxb7 43.Nxa6 [43.Bxa6 Rbb8 44.b7 (44.Rb3 Qxh3) 44...Rxb7 45.Bxb7 Rxa4 46.Rxa4 g6 47.Ra6]

43...Qd7 44.Qc2 For the second time in this game the queen steps out of a pin on the a-file. Now it's a simple matter of dominating the a-file, trading pieces, and pushing the b-pawn. Totally crushing.

44...Kh8 X3D Fritz was still managing to find enough defensive resources in its search to delay the inevitable. Itss evaluation was only -1.50, or a pawn and a half negative when in fact it is completely losing. The final cataclysm is beyond its search horizon, just like the rest of the game.

45.Rb3 1-0 (D6)

The rooks will double on the a-file, penetrate to a7 or a8, force exchanges, and finally the push of the b-pawn will be unstoppable.

[45.Rb3 Ne8 a) 45...Qc8 46.Rba3 g6 47.Nc7 Rxa3 48.Rxa3 Rb8 49.Qa2

b) 45...Qf5 46.Nc7 Rxa2+ 47.Qxa2 Nd7 48.Qa7 (b) 48.Be2) ; 46.Rba3 Nc8 47.Nb4 Rab8 48.Ra8 Bg5 49.Rxb8 Rxb8 50.Ra6 Bd8 51.Qa4 Ne7 52.Ra8 Rxa8 53.Qxa8]


other news

Opening ceremony photos and transcript
Analysis of game one
Analysis of game two
Report on game three from New York City
Report on game two from New York City
Q&A with Garry Kasparov
US Chess Trust | ChessBase