At the press conference, Garry was excited and now had become very confident by the result of his new strategy and his performance. Our programmers, Shay Bushinsky and Amir Ban were startled and shocked by Garry’s successful play. Amir said - “If Kasparov does as well in every game, Junior has no chance.”
I have worked with Kasparov and am very familiar with his game, and was very impressed how well he had adapted to the new competition. Although I was a bit excited with Garry’s strategy, I had not begun to panic.
After all, we were being challenged by the best human player in the world, and I had my hands full making sure there would not be another game 1 accident!
That evening I explained to our programmers Shay Bushinsky and Amir Ban, what Kasparov’s strategy was and in what way I think we had to respond.
The plan was:
1. Reconsider the opening book of the computer.
2. Make serious changes to the program (The blessing, the rules of a match resolved it.)
My problem was, not only let Kasparov increase his advantage , but also to invite him to a different kind of struggle in which it would be hard for him to counteract the program – both in an opening, and in the middle game.
Preparation for each following game required at least 30-40 hours of my time .You see; all of us were more prepared for anticomputer strategy:
I could not believe Kasparov could fight 5 more games to play with the computer in tactical chess. But the fantasy that Kasparov at any moment would back to anticomputer strategy, remained an illusion. We in a match have not seen the Berlin wall or any closed positions. For example, Sicilian with 3.Bb5.
And on a move of a match these problems have left on the third plan.
Except for development of rigid tactics after failure in the 1-st game, it was important for us not to concede the initiative.
The main thing – If we had changed our original Junior common openings in such a situation, there was the great danger that Kasparov would at once would feel this weakness.
(2) Deep Junior – Kasparov,G [B42] NY Match, 2003
1.e4 c5 2.Nf3 e6
Kasparov opts for quite Paulsen variation of the Sicilian defence.
3.d4 cxd4 4.Nxd4 a6 5.Bd3 Bc5 6.Nb3 Ba7 7.c4
Junior takes the center.
7…Nc6 8.Nc3 d6 9.0–0 Nge7 10.Re1 0–0 11.Be3 e5 12.Nd5a5 13.Rc1 a4!
This part of the game Kasparov played very deep. Junior gets into strategically passive position, especially Bd3 feels a bit passive despite Whites space advantage.
14. Bxa7 Rxa7 15.Nd2 Nd4 16.Qh5
Typical Computer move. Human would bring his attention to the Queenside, indeed Junior keeps an eye on the Kingside
Seems like White is standing more actively and Junior is bringing more forces toward the Kingside.
17…Nc5 18.Bc2 Nxd5 19.exd5 g6 20.Qh6 f5 21.Ra3 Qf6?
The critical moment of the game! First Kasparov misses 21…e4! with good counter play.
After 22.Nf3! White maintains an advantage, as b4 is now a serious threat. The problem is that the computer can’t appreciate the amount of compensation black is about to get for the sacrificed material.
22…axb3! 23.Rxa7 bxc2 24.Rc1 e4 25.Rxc2 Qa1+?
A moment of carelessness. Kasparov, a bit tired thinks that everything looks good for him and inserts an unnecessary check. 25…f4! as he himself indicated after the game would have left black with strong initiative.
26.Nf1 f4 27.Ra8!
27…e3 28.fxe3 fxe3 29.Qxf8+
The computer bails out with a draw.
29…Kxf8 30.Rxc8+ Kf7 31.Rc7-c8
is just a perpetual. Black isn’t going to go for more and get less with either Kf6 or Ke8.
The first thing Garry said after the game was that he is happy to finally break the curse of losing game 2 of a match! So now Garry has one win and one ‘almost’ and nearly everyone is already counting the computer out! The gambling site on the Internet that takes bets considers Kasparov as a heavy favourite now.
No doubt Kasparov worked very hard before the match and now tried to use new ideas and patterns in which Deep Junior feels it may be in a trap. (Game 1 is a good example.)
By the way, both matches Kramnik vs Deep Fritz and Kasparov with Deep Junior – developed approximately under one course:
First- domination of the human, then – is escalating on eyes of muscles at the computer. To this there is rather simple explanation.
The human in the beginning of a match always tries to grab the initiative and it a psychological edge. Unfortunately, the computer can not perceive a similar psychology.
The computer doesn’t know anything about the score of the match. The computer doesn’t even know it is playing chess. For the software it is a mathematical problem.
Their main purposes constantly increase the “evaluation”; that is an estimation of a position.
The human, leading in the score, starts to resort to more cautious tactics. It happened both with Kramnik, and with Kasparov.