(36) Comp Deep Blue - Kasparov,G (2785) [B07]
New York man vs machine New York (4), 1997
[Nunn]



1.e4 c6 2.d4 d6
Kasparov sticks to his anti-computer opening strategy.

3.Cf3
After this the game transposes to a type of Pirc. More aggressive responses are 3 c4, which might lead to a sort of King's Indian, or 3 f4.

3...Cf6 4.Cc3 Ag4 5.h3 Ah5 6.Ad3
A slightly passive set-up; it isn't obvious exactly what the bishop is doing on d3. A more dynamic build-up would involve g4, Qe2 and Bg2.

6...e6 7.De2 d5
Black's opening strategy is now clesr. His light-squared bishop has ben developed outside the pawn chain and will sooner or later be exchnaged off. Black then erects a light-squared barricade in the centre, so that if the centre becomes blocked then he will be left with his 'good' bishop. The cost is the tempo wasted in playing ...d6 and then ... d5. However, in the relatively closed positions which result, this tempo is not of great significance.

8.Ag5
This appears to be a new move. Hitherto, White has tried [8.exd5 Axf3 9.Dxf3 cxd5 10.0-0 Cc6 11.Ce2 Ae7 12.c3 0-0 13.Af4 a6 14.Cg3 b5= 1-0 Chernysh ov,K-Maiwald,J/Dresden op 1995 (39) with equality, Chernyshov-Maiwald, Dresden Open 1995. The position resulting from the exchange on d5 closely resemble the line 3 exd5 cxd5 4 Bd3 in the Caro-Kann.]

8...Ae7 9.e5
[Not 9.0-0? which runs into 9...Cxe4 10.Axe7 Cxc3 11.De5 Dxe7 12.Dxg7 Df8 13.Dxf8+ Txf8 14.bxc3 Axf3 15.gxf3 is strategically winning for Black.]

9...Cfd7
[If Black wanted to make sure that his h5-bishop was exchnaged for a knight rather than a bishop, he could haveplayed 9...Axf3 10.Dxf3 Cfd7 11.Axe7 Dxe7 but the problem is that the b8-knight is much harder to develop. Overall, the move played in the game appears most accurate.]

10.Axe7 Dxe7 11.g4 Ag6 12.Axg6
This looks slightly odd to human eyes, because the h8-rook is activated against the backward pawn on h3. It appears more natural to play 0-0-0, then aim to move the knight from f3 and play f4-f5.

12...hxg6 13.h4
However, Deep Blue shows that the exchnage on g6 did have a point. If Black plays a later ...f6, then White can reply exf6, gxf6; g5, undermining Black's control of e5. WE shall see this idea again later.

13...Ca6 14.0-0-0 0-0-0 15.Tdg1
Preparing for kingisde activity by h5.

15...Cc7
A fundamental decision, over which garry thought for a long tiem. Black has two alternative plans: the first is to support the e6-square and then play ...f6. White's likely response is to swap on f6, and then Black can try to play ...e5. If he succeeds in this plan, Black will have a majority of pawns in the centre. The second plan is to attempt a queenside attack. If he were to go for this plan, Black would play 15...Nb6, with the idea of ...Nc4 and ...Qb4 to come. In this case the knight shouild stay on a6 as it might be able to partciipate directly in the attack from this square. The choice in the game shows that Garry has definitely decided for the first plan.

16.Rb1
[The point of this move is not entirely clear - it even looks like a lazy human move, just tucking the king out of harm's way so as to avoid overlooking a possible check later! 16.h5 gxh5 17.gxh5 is more ambitious as the tactical idea 17...f6 18.Ch4 fxe5 19.Cg6 Db4 20.Cxh8 exd4 21.a3 Da5 appears good for White after 22.Cf7 ]

16...f6
Challenging White's central control.

17.exf6 Dxf6!
An excellent and combative move. The obvious continuation was 17...gxf6, but then White could reply 18 g5 eliminating the f6-pawn and thereby preventing Black from playing ...e5. The result would be a general opening up of the kingisde, when White's rooks appear well-placed on h1 and g1. The move played leaves Black with doubled and isolated pawns, supposedly the chess-player's worst nightmare. However, White also has serious structural weaknesses; the pawns on g4 and h4 aren't doing much and, even more seriously, his f3-knight and f2-pawn are vulnerable to pressure down the f-file.

18.Tg3
Deep Blue takes steps to defend the exposed knight, but the rook is clumsily placed here.

18...Tde8
Threatening to break through in the centre by ...e5. This would give Black a clear advantage so Deep Blue immediately prevents it.

19.Te1 Thf8
Very natur al play. The f3-knight is immobilised by the pressure along the f-file, since the f2-pawn is insufficiently defended. Garry has excellent compensation for his slight kingisde pawn weaknesses in the form of active piece-play - just the type of situation in which he normally plays very well.

20.Cd1!
Deep Blue finds an excellent counter to Black's plan. Although this looks passive, the knight wasn;t doing much on c3 and it is much more usefully employed defending the f2-pawn. This frees the f3-knight for action; in some lines White might be able to play Ne5.

20...e5!?
An amazing pawn sacrifice. The logic behind this is that there are only two things wrong with Black's posiiton: the backward e6-pawn and the poorly plaved kngith on c7, which has few active prospcets (the g-pawns are not a significant factor provided Black's pieceds remain active, as White will not have the freedom of action to exploit them). By giving up a pawn, Black solves both these problems at a stroke. The c7-knight can move to e6, and then either to f4 or to c5 and then e4. It is certainly a bold stroke to offer a pawn against a computer, as machines (like Kortschnoj) have the annoying habit of hanging on to material. Very good judgement by Kasparov! [Another factor is that an alternative plan is hard to find. 20...Df4 is met, not by 21.Ce5? Cxe5 22.Dxe5 Dxe5 23.Txe5 Tf4 24.c3 Th8 with strong pressure against White's pawns, but by 21 Qd3 and g6 is hard to defend.; Attempting to activate the c7-knight by 20...Cb5 21.c3 Cd6 fails to 22.Cg5 with Rf3 to come.]

21.dxe5
[After 21.Cxe5 Dxh4 Black has no problems.]

21...Df4 22.a3?
An extremely odd move which has no visible purpose, other than the general one of avoiidng potential back-rank problems. [Neither 22.Dd3 Cxe5 nor; 22.h5 gxh5 23.gxh5 Th8 24.Txg7 Txh5 is promising for White; so the best move is probably 22.De3! Dxe3 23.Txe3 Ce6 ]

22...Ce6
Kasparov gratefull seizes on the free tempo to progress with his plan.

23.Cc3
Deep Blue appears is floundering. This move undoes all the good work that the earlier Nd1 had achieved. 23.Qe3 was still the best move. If White were a human being, one would say that he had 'lost the thead'. It seems odd that computers can have the same problem.

23...Cdc5 24.b4?
A very ugly move. This was clearly the point of White'sprevious move: to be able to expel the knight from c5 without it landing on e4. However, the time White gains is as nothing compared to the serious weakening of his king position. Kasparov must have been mentally rubbing his hands with glee at the sight of this reckless advance.

24...Cd7 25.Dd3 Df7 26.b5
Having played b4 to keep the knights out of c5, it is of course inconsistent to let them back in again. Now White has nothing positive to show for the advance of his b-pawn, while his king position becomes weaker and weaker. White should have just remained passive, although Black has various ways of improving his position, for example by ... Kb8 and ...Rc8, aiming for ...c5. [26.Cd1 Rb8 ]

26...Cdc5 27.De3 Df4
White's collection of pawn weaknesses is now so serious that Black feels no hesitation is swapping off queens into an ending a pawn down! His aim is to reach a spotion with doubled rooks on f4 and f8, when White's whole position will be ceaking at the seams.

28.bxc6 bxc6 29.Td1
Deep Blue struggles to maintain control. This move prepares Nd4 in some lines.

29...Rc7
Allowing the rooks to switch to the b-file.

30.Ra1 Dxe3
A surprising decision, but after considerable analysis I think it is correct. Here Black had the opportunity to play for a direct attack against the white king, which Kasparov might not have tunred down against a human opponent. [However, with accurate defence the upshot is far from clear 30...Tb8 31.Tb1 Txb1+ 32.Rxb1 Tb8+ 33.Rc1 Dc4 34.Cd2 Da6 35.Ce2 also achieves little; 30...Dc4 31.Cd2 (31.Cd4! Cxd4 (31...Tb8 32.Cxe6+ Cxe6 33.Dxa7+ Tb7 34.Da5+ looks good for White) 32.Txd4 Df1+ 33.Cd1 and there is obvious continuation for Black.) 31...Da6 really is dangerous for White]

31.fxe3 Tf7
Doubling rooks will force the f3-knight to move, an then Black's rooks cen penetrate to f2.

32.Th3
Another odd move from Deep Blue. Up to now we could say that Black had a slight advantage, but now White is in deep trouble. [32.Cd4 Tef8 33.Cxe6+ Cxe6 looks better, although Black also has some advantage in this case.]

32...Tef8 33.Cd4 Tf2 34.Tb1 Tg2 35.Cce2
Deep Blue decides to hettison the extra pawn, but there were by now no good alternatives. [35.g5 Tff2 ]

35...Txg4
[It is also tempting to play for the attack by 35...Tff2 36.Cxe6+ Cxe6 37.Cd4 Cxd4 38.exd4 Txc2 39.Tf1 c5 40.dxc5 Rc6 41.Te1 Rxc5 , which also strongly favours Black, but there is no reason to criticise the game continuation.]

36.Cxe6+ Cxe6
All White's remaining pawns are isolated, and Black's pieces are all more active than their White counterparts. With all these advantages, it is incredible that Kasparov failed to win this position.

37.Cd4
Another pawn bites the dust, but this is a reasonable decsiion, because the knight on e6 was exrteremely well-placed and White could hardly avoid shedding one of the weak e-pawns in any case.

37...Cxd4 38.exd4 Txd4 39.Tg1 Tc4 40.Txg6 Txc2 41.Txg7+ Rb6 42.Tb3+ Rc5 43.Txa7
White is once again a pawn up, but his three scattered pawns are far less dangerous than Black's two connected passed pawns. Moreover, Black'sking is activ e wherea White's is trapped on the back rank and subjected to mating threats from the enemy rooks. The position should be winning for Black.

43...Tf1+?
This move effectively loses a vital tempo. The somplest win was by [43...Rc4 44.Tab7 c5 45.Tb2 (45.e6 Te2 ) 45...Txb2 46.Txb2 d4 47.e6 (47.Tc2+ Rb3 48.Rb1 d3 ) 47...d3 48.Tg2 Rc3 and Black's pawns are too strong.]

44.Tb1 Tff2 45.Tb4
Had Kasparov overlooked that this move threatens mate in one? If Black wants to play for a win then he is more or less forced to swap one pair of rooks, but this relieves much of the pressure against White's king.

45...Tc1+ 46.Tb1 Tcc2 47.Tb4 Tc1+ 48.Tb1 Txb1+
Black is eventually forced to exchnage rooks, but in comparison with the note to Black's 43rd move, his king and pawns are further back. The position should probably still be a win, but it is now more trickly.

49.Rxb1 Te2 50.Te7 Th2 51.Th7 Rc4?
[Now the game appears to be a draw. The correct move was 51...d4! After 52.Tc7 (52.e6 Te2 53.e7 Rc4 54.a4 d3 55.a5 Te1+ 56.Rb2 d2 wins) 52...Txh4 53.e6 Te4 54.e7 Rd6 the point is revealed: White loses his e-pawn.]

52.Tc7!
Deep Blue seizes the drawing chance. First of all, the rook is transferred behind the black pawns with gain of tempo. This is necessary, because it is of no value to push e6-e7 while the rook is stuck on h7 - the rook must be available to restrain Black's pawns from behind,

52...c5 53.e6 Txh4 54.e7 Te4 55.a4
This is the second point. Black no longer has the move ...Kd6 available, so White keeps his e-pawn.

55...Rb3
The position is just drawn. With the rook actively placed on c7, Black can only advance his pawns very slowly, too slowly to beat White's a-pawn to the queening square. [55...Rb4 56.a5 ; 55...d4 56.a5 d3 57.a6 ]

56.Rc1 1/2-1/2