Author: Vincent Diepeveen
Date: 04:59:00 07/05/01
Go up one level in this thread
On July 05, 2001 at 05:10:12, Tony Werten wrote: >On July 05, 2001 at 04:58:45, martin fierz wrote: > >>On July 05, 2001 at 03:36:39, Tony Werten wrote: >> >>>On July 04, 2001 at 07:01:55, martin fierz wrote: >>> >>>>On July 04, 2001 at 05:41:22, Dan Andersson wrote: >>>> >>>>>ProbCut and MultiProbCut: >>>>>http://www.neci.nec.com/homepages/mic/publications.html >>>>>AFAIK this is the canonical source. >>>> >>>>thanks! i had missed the multiprobcut paper there. >>>> >>>>has anyone ever tried this in chess? >>> >>>My guess, based on some talking with Ed en Richard, is that Rebel and Genius are >>>using it. ( As a replacement for nullmove, or rather nullmove has replaced >>>probcut ) >>> >>>cheers, >>> >>>Tony >> >>you mean nullmove is better than probcut for chess? or is it 'unclear'? in >>checkers, probcut works fine. > >In checkers nullmove doesn't work. In chess both work, but nowadays everybody >seems to use null. It has some advantages but one big problem: zugzwang. > >Both have the same basic idea: > >Probcut: This position sucks so sever, a shallow search can prove it's not going >to get (much) better, so take alfa cutoff. > >Nullmove: This position is so good, even if I give my opponent another move, a >shallow search can prove it's not going to get worse, so take beta cutoff. > >Make your pick. With nullmove you can return a very correct score, namely the score of the search of the opponent which is >= beta. What score to return with probcut? Because if you return the score of the smaller search, then you're busy with a kind of search which is completely game tree technical spoken going to miss thing after thing, whereas with nullmove the scores you get back at the root are correct. zugzwang you can detect with double nullmove, only a bit bigger depth is usual needed. with probcut you can proof that you sometimes cannot find a move even at depth n if you return the score of the subtree, where n is a huge number but of course not infinite. then there is the huge probcut reduction. usually people use a major reduction factor for probcut. double nullmove is *not* working in probcut. Best regards, Vincent >Tony >> >>cheers >> martin
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.