----- Original Message ---- > From: Rémi Coulom <[EMAIL PROTECTED]>
> According to my experience with Go data, it is not possible to give the > value of one stone in terms of Elo ratings. For weak players, one stone > is a lot less than 100 Elo. For stronger players, it may be more. > > Also, it is very important to understand that the Elo model is very > wrong, and Elo against humans has nothing to do with Elo against > computers (and even less with Elo against the previous version). In > games against GNU Go, Crazy Stone improved 200-300 Elo points in one > year. On KGS, this translated into an improvement from 2k to 1k. Overspecialization? GNU Go has known weaknesses; learn to exploit those, and your win rate can easily approach 100%. But playing against a variety of opponents, with different strengths and weaknesses, one might not do so well. The exploit might even be an objectively weaker play. Many human players rely on "trick plays" which trap unwary opponents; the value of the move depends upon the probability distribution of the opponents' response. The best reply puts you behind, but the more probable replies give you an advantage. In the Mogo vs MyungWan Kim game, Myungwan Kim attempted such an overplay, to which Mogo made the correct reply. As Mr. Kim said, in a handicap game one must make overplays to win - the question is, which overplays will succeed? Winning rates are important, but to really boost their strength, programs should take the time to do post-game analysis to find provably optimal moves which would defeat any conceivable strategy; this knowledge should then be condensed to a form which can be computed more rapidly during the course of actual games. _______________________________________________ computer-go mailing list [email protected] http://www.computer-go.org/mailman/listinfo/computer-go/
