M Winther wrote on Wed, Oct 22, 2008 11:47 AM UTC:
Muller, it's not always relevant to test a program against other programs.
Most chess programs beat 99% of humans. Nevertheless, they are smashed by
Rybka, and completely annihilated. Of course, Zillions loses against
stronger programs. But is it fun and developing to play against it? Yes,
it is. Zillions, on a modern computer, plays chess at, perhaps, Elo
2150-2200. It plays an entertaining game, too, advancing boldly with the
pawns on the wing, etc. I always tweak it to make pawn moves in the
opening, and discourage early queen excursion. Then it plays a rather
mature game. I am a good judge, because I'm a quite strong club player.
Moreover, Zillions understands all those strange new pieces better than
humans do. So it often wins by utilizing them better. It takes longer time for a human to develop a grasp of a cannon or a catapult, etc. Zillions is clearly the best way of testing chess variants, whether they are fun, whether they are interesting strategically, etc. Zillions is a remarkable tool for chess variants. People in the chess community are often interested in which program has the highest rating, but who cares? Zillions is actually the best chess program because it plays most variants.
/Mats