On Dec 5, 2007 11:33 AM, Erik van der Werf <[EMAIL PROTECTED]> wrote:
> On Dec 5, 2007 5:01 PM, Álvaro Begué <[EMAIL PROTECTED]> wrote: > > On Dec 5, 2007 9:33 AM, Erik van der Werf <[EMAIL PROTECTED]> > wrote: > > > > > > Look for Realization Probability Search. > > > > Oh, thanks! I knew it was too natural to be original. Well, I actually > > thought about it around 1998, so it might have been new back then. > > Maybe, but the closely related idea of adding fractional plies for > search extensions has been around quite long. I would not be surprised > if some Chess engines already used it in the 90's (but of course > nobody would talk about it). Oh, I am sure a lot of people used fractional plies for chess in the 90s, including myself. There was even a paper in the ICCA journal in 1989 titled "The SEX algorithm in Computer Chess" which described a similar method, but the rules for deciding how many fractional plies should be subtracted from the depth were never linked to an actual probability distribution, and combinatorial explosion of certain branches of the tree was possible (while you can't have that happen with Realization Probability Search, since no more than 1,000,000 nodes can have probabilities over 0.000001 on a given level of the tree). Anyway, whether I discovered this before other people or not is irrelevant; as I said the idea is very natural. The point is that this is probably the right framework when you are trying to think of how to organize a search that will find everything given enough time and still allows for selectively exploring promising moves when time is limited. Or at least it's an approach that people should consider. Álvaro.
_______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
