When Deep Blue lost the first game of its match against Garry Kasparov, it looked as though its programmers had not overcome the basic defects exposed in their first encounter last year. The machine was still making strategic errors in its search for tactical advantage. In the second game, however, the 200 million moves a second supercomputer showed us all that perhaps it does know what it is doing after all. Indeed, it outplayed the world champion so comprehensively that we must ask ourselves what it was playing at in the first game.

Kasparov's opening was an anti-computer special designed to keep the two sides pieces apart and avoid any complications. After the game, he took some pride in pointing out that he had won without ever moving anything more valuable than a pawn into his opponent's side of the board.

The fun started with 13.Nh4 when the threat of g4 followed by Nxg6 induced Deep Blue to play the positionally weakening 13...g5. When 22...g4 followed, Black seemed to be attacking, but in reality was only adding to his problems. Perhaps the machine had already seen the idea of 25...Be7 and 26...Bg5, followed by 28...f5 and 29...e4. It was all very ingenious, and led to the win of rook for bishop and pawn, but the strength of the white pawns that were left on f5 and g6 was always going to win the game for Kasparov.

Whatever one may think about its positional play, Deep Blue certainly made Kasparov work for his win.

White: Garry Kasparov

Black: Deep Blue

Game One, New York 1997

1 Nf3 d5 24 f3 Nxe3

2 g3 Bg4 25 Nxe3 Be7

3 b3 Nd7 26 Kh1 Bg5

4 Bb2 e6 27 Re2 a4

5 Bg2 Ngf6 28 b4 f5

6 0-0 c6 29 exf5 e4

7 d3 Bd6 30 f4 Bxe2

8 Nbd2 0-0 31 fxg5 Ne5

9 h3 Bh5 32 g6 Bf3

10 e3 h6 33 Bc3 Qb5

11 Qe1 Qa5 34 Qf1 Qxf1+

12 a3 Bc7 35 Rxf1 h5

13 Nh4 g5 36 Kg1 Kf8

14 Nhf3 e5 37 Bh3 b5

15 e4 Rfe8 38 Kf2 Kg7

16 Nh2 Qb6 39 g4 Kh6

17 Qc1 a5 40 Rg1 hxg4

18 Re1 Bd6 41 Bxg4 Bxg4

19 Ndf1 dxe4 42 Nxg4+ Nxg4

20 dxe4 Bc5 43 Rxg4 Rd5

21 Ne3 Rad8 44 f6 Rd1

22 Nhf1 g4 45 g7 resigns

23 hxg4 Nxg4