>If the utility of any win is the same, it makes sense to simply maximize the probability of winning. If we are not happy with the program wasting points in a favorable endgame, it must be the case that we are happier with a win by a large margin than with a win by a small margin
I don't see a problem with entering a different regime, for the purpose of playing elegant moves in the endgame, once we have already secured a win. Our engine's strength will not suffer. > Third, the "only wins matter" approach seems to discard a great deal of > useful information. > Yes, it does seem so. Now only if we could show any information besides all-or-nothing result to be useful :) I have thought about looking at this a bit more systematically and framing it like a machine learning problem: learn what features of a playout give you information about the correct evaluation of the position. Has anyone tried this? If some playouts really should count more than others, then it might take this kind of approach to decide exactly how to count them.
_______________________________________________ Computer-go mailing list Computer-go@dvandva.org http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
