The Chess Mind

Running alongside the (human) world championship, the computer programs Rybka and Zappa played a match for silicon supremacy. To the surprise of most, Zappa was the winner, 5.5-4.5, though there were a couple of unusual aspects to the win. I referred in the previous post to Kramnik's super-deep preparation against Anand in their first game, but that only went as far as move 28. In game 6, Zappa's theoretical preparation went to move 37, where it calculated the position to the perpetual check that ended the game on move 52. Niiiice. Is that really the program's half a point?

And then there was the bizarre fourth game. In a position with with a queen and five pawns for Rybka against Zappa's two rooks, a bishop and a pawn, White had the ostensible advantage but Zappa an unbreachable blockade. When it was clear that Rybka couldn't break through, its operator offered the Zappa team a draw. The Zappers refused, not based on the program's evaluation but on computer "psychology". They reasoned, rightly, that Rybka might (a) think its position was still better even without one of those pawns, and that (b) it might therefore chuck that pawn to avoid a 50-move rule draw. Sure enough, that's what happened: Rybka thought it was still better, gave up the pawn for the reason given, and then went on to lose.

In any case, it was an impressive success for Zappa, demonstrating that Rybka isn't the only big dog on the block. (Actually, they're all big dogs as far as we humans are concerned, but if it's important that your program play at a 3000 level rather than a crummy low-2900 status, you now have a second option to consider.)

Links here and here.