X3D Fritz - Garry Kasparov (2830) [D27]
X3D Man-Machine World Championship
New York (4), 18.11.2003

1.d4 Already a bit of a surprise from X3D Fritz. Computers almost always play 1.e4 against humans because it leads to open positions with sharper, more concrete play, at least speaking very broadly. Of course any chess opening can lead to a sharp position, as we see here. The X3D Fritz team had a specific opening line in mind when they went for 1.d4 in this critical game four.

1...d5 For a long time Kasparov played 1...Nf6 exclusively against 1.d4. That move usually led to the aggressive Grunfeld or King's Indian Defenses. Lately, however, he has been playing the more solid move here, staking out space in the center right from the start.

After the game Kasparov said that he wasn't completely surprised by 1.d4 and had done some preparation with his seconds for the match, Grandmasters Yuri Dokhoian and Mikhail Kobalia. (Dokhoian is his long-time trainer.)

2.c4 The Queen's Gambit, offering a pawn to deflect Black's central d-pawn from its solid post. This has been one of the most common opening systems at every level of the game.

2...dxc4 The Queen's Gambit Accepted, as Black accepts the offer of the pawn. Unlike other gambits Black only rarely attempts to hold on to his extra pawn in this variant. Doing so subjects him to too much attacking pressure. Instead he develops his pieces. Kasparov has played this before but the commentators weren't sure he would enter such an open, double-edged opening against X3D Fritz.

3.Nf3 e6 4.e3 Nf6 5.Bxc4 Reestablishing material equality. 5...c5 Black counterattacks the white center immediately. 6.0-0 a6 Threatening to play 7...b5 and 8...c4 with a pawn advantage on the queenside.

7.Bb3 [7.dxc5 This is a common line, but of course the X3D Fritz team wouldn't want to see the queens come off the board like this. They want as much activity and piece power on the board as possible to press the computer's advantage in tactics and calculation. 7...Qxd1 8.Rxd1 Bxc5; 7.a4 Nc6 8.Qe2 cxd4 9.Rd1 Be7 10.exd4] 7...cxd4 (D1)

Kasparov heads into a line he has played before. You can see why the X3D Fritz team liked this line. All the pieces are still on the board and there are open lines for attack and piece play. There are no lines of blocked pawns to inhibit the computer's calculating ability.

The downside is unrelated to the objective characteristics of the position. It's that Kasparov knows this position very well and is likely to have studied it deeply. This turns out to have been exactly the case.

[7...b5 This move, quickly developing the bishop to b7, is the most popular here. 8.a4 b4 9.Nbd2 Bb7]

8.exd4 Now White has what we call an isolated queen's pawn, a very common structure. The pawn on d4 has no pawns alongside it to defend it and it can be a target for Black's forces. In compensation White has excellent open lines for developing his pieces and the pawn controls important center squares. It is also possible to advance the pawn to start an attack.

8...Nc6 9.Nc3 Be7 10.Re1 0-0 11.Bf4 This move made it clear to Kasparov that the X3D Fritz team was headed into a line that was played in several very high-profile games a few years ago. Kasparov played in two of them, one with white in 1999 and one with Black in 2001!

That 2001 game was against no lesser an opponent than world #2 Vladimir Kramnik on the stage in Moscow. Kasparov had won that game and no one had dared to play that line for White since then. Kasparov now started to play slower, warily looking ahead for what X3D Fritz had in mind.

More precisely, what the X3D opening book team had in mind. The machine was still playing out of its opening library, not thinking on its own at all. There is a dead giveaway by the way the machine plays its moves instantaneously when it is still in its database book. When that's going on Kasparov knows that he is still following the programmers' preparation.

11...Na5 Attacking the bishop. Exchanges are usually to the defender's benefit, especially in a cramped position. [11...b5 12.d5 exd5 13.Nxd5+/- Nxd5 14.Qxd5 Bb7 15.Qh5 Bf6 16.Rad1 Qc8 17.Bd6 g6 18.Qh6+- 1-0 Epishin,V-Jonkman,H/Amsterdam NED 2000/(30)]

12.d5 Wantonly giving up the weak pawn in order to open lines and create complications. Kasparov himself played 12.Bc2 against world #3 Anand in 1999.

[12.Bc2 b5 13.d5 1/2-1/2 Kasparov,G-Anand,V/Wijk aan Zee 1999/CBM 69/[Huzman] (22) (13.Qd3 Bb7 14.Be5 g6 15.Qe3 Nc4 16.Qh6 Bxf3 17.gxf3 Nh5 18.Ne4 f5-+ 0-1 Akesson,R-Degerman,L/Ronneby 1998/EXT 99 (23)) ]

12...Nxb3 13.Qxb3 (D2)
 

After this Kasparov went into a very long think. He had faced this exact position two years ago in Moscow against his arch-rival, world #2 Kramnik. They were playing a match of blitz, games with just five minutes per player instead of the two hours plus of classical chess.

Kasparov's was wondering if he should continue to follow the line from that game. He beat Kramnik spectacularly in that 2001 blitz game by sacrificing his queen and winning a wild tactical melee.

But surely trying that against X3D Fritz would be close to suicide. It was just that sort of position the computer team dreamed of when they told X3D Fritz to play this line if given the chance.

So Kasparov sat there for a while looking at the alternatives. He was mentally thumbing through his years of memorized analysis and looking to see what dangerous improvements the X3D Fritz team might have found.

13...exd5 Kasparov avoids the wild queen sacrifice line and recaptures with the pawn. Grandmaster Yasser Seirawan, one of the ESPN commentators, didn't like this line at all against a computer because of the many open lines.

The alternatives were worse, however, at least according to Kasparov. This move has also been played before in top level games and this is a factor in Kasparov's favor. He would definitely have analyzed those games.

[13...Nxd5 A remarkable queen sacrifice originating with Alexei Shirov. 14.Rad1 Nxf4 15.Rxd8 Rxd8 16.Rd1 Nd5 17.Ne5 (17.Nxd5 exd5 18.h3 b5 19.Rxd5 Be6 20.Rxd8+ Rxd8 21.Qc2 1/2-1/2 Gelfand,B-Shirov,A/New Delhi/2000/CBM 80 (21)') 17...Bf6 18.Nc4 Rb8! (18...Nxc3 19.Rxd8+ Bxd8 20.Nb6!) 19.Na5?! a) 19.Nxd5 exd5 20.Ne3 Be6 (a) 20...d4 21.Nd5 Be5 22.Re1 Bd6=) 21.Nxd5 Bxb2 22.Qxb2 Rxd5 23.Rxd5 Bxd5=

b) 19.Nb6 Nxb6 20.Qxb6 Bd7 21.Ne4 Be7= /=/+; 19...Bd7 20.Ne4 (20.Nxd5 exd5 21.Rxd5 b6 22.Nc4 Be6-/+) 20...Be7 21.Nc4 Bb5-/+ 22.Ne5 Be8 23.h3 b5 24.Qg3 Rbc8 25.Kh1 a5 26.h4 a4 27.Ng5 h6 28.Ngf3 Nf6 29.Rxd8 Rxd8 30.Qf4 Rd1+ 31.Kh2 Bd6-+ 32.g3 Bc6! 33.g4 Rd3! 34.Kg1 Bxf3 0-1 Kramnik,V-Kasparov,G/Moscow 2001]

14.Rad1 Played instantly, showing that X3D Fritz was still in its opening library. Black is unable to hold on to the extra pawn.

14...Be6 15.Qxb7 Getting the pawn back, but now the queen is exposed on b7. 15...Bd6 Kasparov offers exchanges to weaken White's attacking forces. Another move played here is the much more complicated 15...Bc5.

Anand played that move against Kramnik in 2001 and a few months later his Indian compatriot, Sasikiran, tried it against Bacrot. The extreme complications arising from that move are exactly what any human would want to avoid against a computer beast like X3D Fritz.

[15...Bc5 16.Be5 Qa5 (16...a5 17.Nd4 Nd7 18.Bg3 Re8 19.Ndb5 h5 20.Nxd5 h4 21.Bf4 Bg4 22.Nbc7 Rxe1+ 23.Rxe1 Rb8 24.Qxb8 Nxb8 25.Re8+ Qxe8 26.Nxe8 Nc6 27.Nef6+ gxf6 28.Nxf6+ Kf8 29.Nxg4 1-0 Bacrot,E-Sasikiran,K/Lausanne 2001/CBM 82 ext (42)) 17.Nd4 Bxd4 18.Rxd4 Nd7 19.Bd6 Nc5 20.Qc7 Qxc7 21.Bxc7 Rfc8 22.Bg3 Ne4 23.Nxd5 Bxd5 24.Rxd5 Nxg3 25.hxg3 1/2-1/2 Kramnik,V-Anand,V/Monte Carlo 2001/CBM 81 ext (76)]

16.Bg5 Keeping the bishop on the board and pinning the knight on f6 against the queen. 16...Rb8

17.Qxa6 Kasparov was relieved when the computer actually thought for a while before playing this obvious move. That meant it was finally out of its opening library and he wasn't out of his!

Now he didn't have to worry about a nasty prepared surprise. Kasparov had analyzed this position extensively in the past and didn't think Black had anything to fear. 17...Rxb2 (D3)
 

18.Bxf6 Making things easier for Kasparov. The world #1 had already seen through to the completely equal endgame that is now coming.

X3D Fritz could have kept more material on the board instead of exchanging, but it doesn't know how to do anything other than play the best move.

If a move that simplifies is rated as 0.01 points better than an incredibly complicated move it will play the simplifying move even though it would have much better winning chances against a human by playing the complicating move.

One thing the X3D Fritz team might not have been aware of is that the previous game to reach this position was played by GM Andrei Kharlov, a Russian who has been part of Kasparov's analysis team several times. This would significantly increase the chance that Kasparov would know this line inside and out.

[18.Re2 Rxe2 19.Qxe2 Be7 20.Qd3 Qa5 21.Bxf6 Bxf6 22.Nxd5 Qxa2 23.Nxf6+ gxf6 24.h3 Qa5 25.Qd4 Kg7 26.Qe3 Re8 27.Rc1 Rc8 28.Rxc8 1/2-1/2 Dorfman,J-Kharlov,A/France 2001/EXT 2002 (28)]

18...Qxf6 Kasparov was now playing his moves quickly and with a peaceful look on his face. He knew the danger was past. Black has absolutely no winning chances in these lines as there is no way X3D Fritz is going to make a huge blunder. In simple, open positions computers are all but invincible. The draw was now becoming apparent.

19.Qxd6 Qxc3 20.Nd4 If White tries to save the a-pawn it can come out on the worse end of the position. X3D Fritz is content to play a line in which it eliminates Black's central passed pawn. [20.a4 Ra2]

20...Rxa2 21.Nxe6 fxe6 22.Qxe6+ Kh8 It's White to play, but Black's double attack on the f2 pawn forces X3D Fritz to defend instead of grabbing the d5 pawn. It's completely equal with no winning chances for either side.

23.Rf1 [23.Qf7 Qc8 24.Qxd5 Raxf2 25.Rf1 Qc2 26.h3 h6] 23...Qc5! Continued precise play from Kasparov. The triple attack on f2 leads to further exchanges. [23...Rd2 24.Qe1 Rd3 25.Qxc3 Rxc3 26.Rxd5]

24.Qxd5 Rfxf2 (D4)
 

The point! Black isn't losing his queen here because he is threatening checkmate in two moves. Of course X3D Fritz had seen all this coming too. Still, it's a pretty way to end a rather technical final game.

Both sides have serious back-rank problems; their kings have no way to escape a check on the rank. With all the heavy pieces on the board this creates various tricks, but both Kasparov and X3D Fritz were up to the task.

25.Rxf2

[25.Qxc5?? Rxg2+ 26.Kh1 Rxh2+ 27.Kg1 Rag2# Checkmate!; 25.Qd8+ Rf8+ Blocking check with check, a rare and attractive tactical theme that saves the day. Kasparov had foreseen all of this as early as a dozen moves ago. 26.Kh1 Raf2 (26...Rxd8?? 27.Rxd8+ Qf8 28.Rdxf8#) 27.Rxf2 Qxf2 28.h3]

25...Qxf2+ 26.Kh1 h6 (D5)
 

Eliminating the back-rank weakness by giving the king some air (usually called by the German "luft").

10 moves earlier online commentator Mig Greengard has predicted the game would finish drawn in this position. But X3D Fritz ruins that prediction by playing one more useless move first.

27.Qd8+

[27.Rc1 Any human would have played this move just to see if Black might blunder. It isn't immediately obvious how Black can deal with the threat of Rc8+ with a winning attack. The black rook can't go to a8 and if it blocks with ..Rc2 then Qd8+ Kh7 Qd3+ wins the rook with a fork. Three-time US Champion Joel Benjamin was momentarily alarmed, thinking that perhaps Kasparov had missed this trick in what looked like a safe position at last.

27...Qc2! This is the way! Black offers the queen to block the check and White cannot capture without being checkmated! A nice trick that Kasparov had to have seen long ago or this line could have meant trouble. It turns out there is another way to save Black, but it's not nearly so clear or pretty. (27...Ra6?! Unnecessarily complicated. 28.Rc8+ Kh7 29.Qe4+ Rg6 30.h4 Qf1+ 31.Kh2 Qf6 32.h5? Qd6+ 33.Kh1 Qd1+ 34.Kh2 Qxh5+) 28.Rxc2?? (28.Rf1 Kh7) 28...Ra1+ 29.Rc1 Rxc1+ 30.Qd1 Rxd1#; 27.Rb1 Rb2 28.Ra1 Rb8]

27...Kh7 White cannot make progress and has to worry about its own king and the attack on g2. 1/2-1/2