what i'm saying is that monte carlo is not evaluating the game-theoretical value of the board. what it is doing is looking for best moves with respect to the function: "maximize probability of win". probability of win is not zero at the start, even with 6h. it is lower than without handi, but not zero. to see this, imagine a zero komi game where you play as white. MC should still play decent moves in this case.
neither player plays anything close to game-theoretic best moves in any games -- if they did, or even had an evaluation function that was capable of making such an evaluation, they could quite easily beat any professional player. i think that the main objection to this is really that they don't have an opening library in a handicap game. in reality, opening libraries don't make much sense in go, unless you can pick josekis that work together from opposite sides of the board to satisfy some larger goal, which would not happen in general with an MC program that didn't know about influence, etc. s. __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
