On Mon, 2008-01-14 at 19:41 -0800, mingwu wrote: > 1K ~ 100 K / sec is much faster than "a dozen" / sec of a conventional > program. > > Do they calculate dragon safety (eyes, connections, patterns ...)? if > not, the estimate will be VERY unreliable.
That's just it, they don't. They play a *random* game and get a win or loss (0 or 1) out of it. Random games in better positions tend to win more frequently. > reevaluating positions more than once (maybe 100 times?) to > get a more > reliable estimate. > > > > why "reevaluating" the same position? It's like flipping a weighted coin... It takes many flips to figure out how frequently it comes up heads or tails. > Sorry, I didn't go into their papers, can people who knows UCT, or > actually working on UCT programs explain in a way that a layman can > understand. Thanks. That's really the basics of how monte carlo go works. UCT is a method of choosing which coins to flip and how to combine results deep in the search tree. I'm sure many of us are surprised how well this stuff works. _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/