Author: Vincent Diepeveen
Date: 07:41:51 10/28/02
Go up one level in this thread
On October 27, 2002 at 16:49:19, Dan Andersson wrote: Nullmove works great for: chess Go Shogi Chinese chess And a big number of board games where no progress is no good. It works for real life! In real life no progress is bad! So your games i never heard of, perhaps cannot use nullmove. Bad luck for you. Depth limited Alfabeta with Nullmove is definitely A* algorithm, because you basically look at every path. You do not *miss* some lines completely which happens in B* where you directly get thrown in the deep and then hopefully look in the right game space. It is proven to work great for chess the nullmove. For Go it will take a few years, because the nullmove basically starts to work well when you can get a decent depth, which in GO they do not get yet. There is no legal null move in Chess either, where in GO it even is legal and the worst move until you stop the game. It is not something that assumes that the pass option is a legal move. It is a method to get rid of loads of crap lines! Just like the alfabeta pruning is a correct form of pruning, the double nullmove technique with reduction factor is that too. In general majority of GO programs run years behind on chess programs. That is logical. However, it is a matter of time always. The professional go programmers very clearly check out what happens in the computerchess world a lot. They better do :) It is years 80 logic to say that reducing the branching factor with a correct pruning technique is a bad thing to do :) >You said nullmove works great for most games, and it is not so. Not games with >players numbering over x worldwide players. So I fail to see the point in what >you write. Misere games tend to have fewer players mainly because they are >harder more complex games than non misere games. As well as being variants of >popular games. And we lazy primates seem to prefer positive formulations and >goals on top of that. > I understand your use of algorithm, but to call it null-move is plain wrong. >Even if it evokes a basic analogy. It's plain pruning since there is no null >(void) move in Go. Your use of technical terms is careless as usual. What you >are doing is akin to B*, where the pass is the pessimistic score in most cases. >But the depth of the search is variable in your algorithm, that's not the case >of B*. That might be an improvement or at least an easy way to fold it into A-B >search. What concerns me is what the consequences of recursive application of >the algorithm might have. But I do think that it might be a valid try for a go >playing algorithm. What kind of scoring algorithm do you use in the leaf nodes? >Is it exact (unlikely as it seems) or merely approximate. And what depth are you >attaining? > >MvH Dan Andersson
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.