banner IBM
toolbar
May 7, 1997

Kasparov Draws Third Game Against Deep Blue

By BRUCE WEBER

NEW YORK -- Garry Kasparov, the world chess champion, opened the third game of his six-game match against the I.B.M. computer Deep Blue Tuesday in peculiar fashion, by moving his queen's pawn forward a single square. Huh?



Related Article
Chess Column: After Another Cautious Start, a Small Gamble Doesn't Help
"I think we have a new opening move," said Yasser Seirawan, a grandmaster providing live commentary on the match. "What should we call it?"

Mike Valvo, an international master who is a commentator, said, "The computer has caused Garry to act in strange ways."

Indeed it has. Kasparov, who swiftly became more conventional and subtle in his play, went on to a draw with Deep Blue, leaving the score of Man vs. Machine at 1 1/2 apiece. (A draw is worth half a point to each player.) But it is clear that after his loss in Game 2 on Sunday, in which he resigned after 45 moves, Kasparov does not yet have a handle on Deep Blue's predilections, and that he is still struggling to elicit them.

He even went so far as to suggest that the computer's decisions in Game 2 were unfairly supplemented by human beings.

His play Tuesday with the white pieces, cautious rather than aggressive, was uncharacteristic, at least compared with games against living, breathing opponents.

"He's not playing his confident, dominant style," Valvo said.

"I don't want to say he's afraid," said Miguel Illescas, a Spanish grandmaster who is with the Deep Blue team. "But when the world champion with the white pieces doesn't want to attack, what do you do?"

Kasparov had resigned the second game on Sunday, and at first the chess world agreed that he had a lost position. But then, a chess player in Michigan found a sequence that the champion (and the computer) had overlooked, and further study showed that had he played that line, he could have forced a draw.

Not only would that have left Kasparov with a 1* to * lead in the match, the very fact of a champion's resigning a drawn position in so important a game is nearly unprecedented.

"I asked him if it had ever happened to him before, and he said no," said Frederick Friedel, Kasparov's computer consultant.

Seirawan added that psychologically, it had to be a devastating blow. "It's unthinkable," he said. "I don't see how Garry could even sit down to play today."

Seirawan is not Kasparov, who is known for his resilience and psychological wherewithal. Friedel said that after the intial shock, Kasparov returned to concentrating on the match. Asked if he was fighting mad, his usual response to a loss, Friedel said no. He was curious, interested and determined.

"We still don't understand how it plays," Friedel said. "We want to find out what are the parameters, how it could play so well and then so badly."

On Sunday, when Kasparov stood and offered his hand to the IBM scientist across the board, the audience, which included a couple of dozen grandmasters, was shocked. But the reaction was less to the resignation than to the demonstrated strength of the computer's play. No one doubted the champion's decision.

The line of play that would have drawn apparently surfaced in an Internet chat room devoted to the match. Credit for discovering the key move for black, Rook to E8, has been claimed by Tim McGrew, a philosophy professor at Western Michigan University who was discussing the match as a member of the Internet Chess Club.

"It was just a flash of insight," McGrew said. "The process of verifying that it does, in all variations, lead to a draw was a team effort."

The strategy involved creating what is called perpetual check, in which Black could keep on checking the white king, move after move, ad infinitum.

"He let his nerves control him," said Patrick Wolff, a grandmaster who was struggling to explain Kasparov's lapse. "What we saw was Kasparov lose control of his nerve."

Other grandmasters felt differently. Kasparov didn't see the drawing strategy, said David Levy, an international master, because he felt the computer would have insured that a perpetual check was impossible. That was seconded by Friedel, who said Kasparov told him, "The computer had played so well I didn't even consider it."

After Tuesday's game, Kasparov concurred, saying the amazing thing to him was that after playing a brilliant game, the computer erred on its last move, allowing the possibility of perpetual check.

"Anyone who knows chess, and a little bit about computers, knows there was a big difference between Games 1 and 3 and Game 2," he said.

"Today the computer was a computer. Sunday something was completely different. Something truly unbelievable happened, and it it showed a sign of intelligence.

"I don't know how it happened. But the most amazing thing is that computer made a blunder on the last move. Suddenly the machine missed an elementary draw."

Was he suggesting that there had been some human interference?

In response, he compared the game to a famous World Cup soccer match in 1986, in which the Argentinian star Diego Maradona fisted the ball into the net for a decisive goal against England, an illegal play that went unremarked by the officials. "Maradona called it the hand of God," Kasparov said.

Murray Campbell, an IBM researcher on the Deep Blue team, shrugged off Kasparov's sinister suggestion. "He can't be happy," Campbell said, "particularly after making such a good start. He doesn't know how we did what we did, and at the end of the match, we'll tell him."

As for the perpetual check, Campbell admitted, "Deep Blue missed it."

"Yes, it was a perpetual check," Campbell said. "But it turned out it was a very deep perpetual check, at least 15 moves down the line." In other words, it was beyond the computer's search, as it was apparently beyond the intuitive powers of the champion.

But it was not beyond the collective powers of a Midwestern professor and the other chess amateurs who were in the same chat room.

The drawing strategy was forwarded to Friedel late Sunday night by a friend in Germany. Too tired to work it out, he put it off until the next morning. Enlisting the help of Kasparov's second, Yuri Dokhoyan, and their own sophisticated computer, known as Fritz, the two analyzed the line and discovered the awful truth.

"Then we debated, 'Are we going to tell him?' Friedel said. "And we decided yes, because otherwise, the first taxi driver we met would tell him. The second question was whether to do it before or after lunch."

It was Dokhoyan who broke the news to Kasparov as the three men walked along Fifth Avenue to lunch at an Italian restaurant.

"Garry stopped in his tracks, and grabbed his head," Friedel said. "There was no shouting, no obscenities. He eventually walked on, didn't say anything.

"At the restaurant, I could see he was analyzing in his head -- click-click-click, click-click-click -- and after five minutes, he glanced at me. And he said, 'Rook e8, h4, h5. That was all? That simple?' How could the computer not see that?"'


Related Site
The following link will take you to a site that is not part of The New York Times on the Web, and The Times has no control over its content or availability. When you have finished visiting this site, you will be able to return to this page by clicking on your Web browser's "Back" button or icon until this page reappears.


Home | Sections | Contents | Search | Forums | Help

Copyright 1997 The New York Times Company



IBM