> Really? You think that doing 20-50 uniform random playouts and > estimating the win probability, when used as a leaf node evaluator in > tree search, will outperform anything else that uses same amount of > time?
Same amount of clock time for the whole game. E.g. if playing 20 random playouts to the end of the game takes N ms, and your static evaluation takes 5*N ms then the playout version can explore 5 times more global nodes. But the main advantage, I feel, is that playouts play out the game. Here is an observation that I think I wrote here many years ago: Many Faces (version 9 or 10) would give a much more accurate analysis of the score of a game by putting it on level 1 and having it play it itself, which took about a second (even with all the overhead of the moves being shown on the board), compared to pressing F9 to get the static evaluation. As already mentioned, life/death analysis is so hard that it just has to be played out. But also, the value of having sente can realistically only be discovered by playing out. This really stumped me when I was trying to statically analyze endgame positions (i.e. trying to extend the Mathematical endgame ideas back a bit further in the game). Darren -- Darren Cook, Software Researcher/Developer http://dcook.org/mlsn/ (English-Japanese-German-Chinese-Arabic open source dictionary/semantic network) http://dcook.org/work/ (About me and my work) http://dcook.org/blogs.html (My blogs and articles) _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/