The big question a computer cannot ask: Is Kasparov making the wrong moves?

What is Garry Kasparov playing at in New York? After four games of his chess match against the IBM computer, Deep Blue, he has scored a win, a loss and two draws. He should have drawn the game he lost - indeed, his game could still have been saved in the final position when when he resigned - and he held a clear advantage in both drawn games. At the end of the fourth game yesterday, Kasparov said: "I believe it was a win ... but I was tired and I couldn't figure it out."

Is this really the man who has spread terror throughout the chess world for the past 12 years? Did Garry Kasparov ever resign a drawn position against Anatoly Karpov? Was he ever too tired to figure out how to beat Nigel Short? Two theories have emerged to explain Kasparov's unconvincing form in the match so far.

The rational explanation is that the world champion is finding it extremely difficult to adjust to the unfamiliar problems posed by a computer opponent. Human beings tend to miscalculate in tactically complex positions. When a position explodes into a sudden turmoil of possible captures, checks and brutal threats, the human mind turns to mush. But that is exactly where computers are at their strongest.

It is relatively simple for a programmer to instruct his machine to look at every possible sequence of captures and checks, and thinking at some 250 million moves a second, you can be fairly sure that Deep Blue will have sorted out all the tactics by the time Kasparov has written down his previous move on the score sheet.

When there are no tactical opportunities, however - in blocked positions for example, or where the White and Black pieces have yet to meet in hand- to-hand combat - computers can only stumble through billions of possibilities, hoping that something good turns up.

In both the third and fourth games of this match, Deep Blue played some atrociously pointless, or gratuitously weakening moves when it did not understand what was happening, but once Kasparov developed direct threats, it found the perfect defences to counter them. Humans, when they start playing badly, generally have the good grace to continue doing so. Facing such a mixture of mindlessness and perfection is enough to throw anyone off balance.

But for anyone determined to reject so logical an explanation, there is a bizarre theory that has been advocated by some suspicious chess followers in the darker corners of the Internet. Just suppose a world chess champion was challenged to a $1m match by a large computer company. If he gave it too sound a thrashing, he would win the prize, but it would not come back for more.

The result of last year's match between Kasparov and Deep Blue turned out to be perfect for all concerned. Deep Blue won one game, which led to unparalleled publicity for its makers as well as encouraging them to believe their programming was on the right lines. There were two draws in the middle of the match, which sustained interest to the very end, then Kasparov ran away at the end with two contemptuously easy victories. The perfect plot. And the first four games this time have followed an identical pattern.

Has Kasparov been pulling his punches? Is he just toying with Deep Blue, in the knowledge that he can dispatch it to the junk yard whenever he chooses?

No, of course he isn't. But if he wins the last two games of the match, proponents of the alternative theory will take it as strong evidence for their case.

But there is a far stronger piece of evidence pointing to the opposite conclusion: if the expression on Kasparov's face was anything other than genuine anguish while he was being pushed around by Deep Blue in the second game, then he is an even better actor than he is a chessplayer. He is, on his usual form, a good enough player to polish off Deep Blue this weekend and take the $700,000 winner's purse. And next year IBM will surely be back for more.