On Dec 18, 2007 3:12 PM, Don Dailey <[EMAIL PROTECTED]> wrote: > > > Harald Korneliussen wrote: > > Some thinking out loud here on the topic of languages and efficiency: > > > > I'd like to know how well MoGo would have played if you let it think > > for a week for every move. Only it seems to me that is not possible, > > because I don't think MoGo will run for a week without crashing. > > Crazystone also crashes quite a lot, if I understand the comments in > > KGS logs correctly. > > > Memory is an issue with these programs, since they build tree's and > maintain them in memory. So none of these programs can think for more > than a few minutes per move.
dimwit gets around this problem by increasing the number of visits required before a node is added to the UCT structure as the structure gets big. This works fairly well, and I can let it run for hours on a position. However, the standard boost of the uncertainty in the score by sqrt(log(total_visits)) grows way too slow, and the main line goes way too deep, without giving other moves a fair chance. I suspect that for very long time controls we would be better off turning UCT (with, say 10K playouts) into an evaluation function and then using alpha-beta on top of it. Álvaro.
_______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
