ChessUSA.com TalkChess.com
Hosted by Your Move Chess & Games
 
 FAQFAQ   SearchSearch   MemberlistMemberlist   UsergroupsUsergroups   RegisterRegister 
 ProfileProfile   Log in to check your private messagesLog in to check your private messages   Log inLog in 

A Go algorithm for chess programmers to try ?
Post new topic    TalkChess.com Forum Index -> Computer Chess Club: Programming and Technical Discussions Flat
View previous topic :: View next topic  
Author Message
Rémi Coulom



Joined: 24 Apr 2006
Posts: 350
PostPost subject: Re: A Go algorithm for chess programmers to try ?    Posted: Wed Jun 06, 2007 6:21 pm Reply to topic Reply with quote
Gerd Isenberg wrote:
that sounds very interesting and promising - you somehow autoscale the weights of different aspects of a move to get a score for move-sorting and/or reduction/extension/pruning. But I still have to think how to differentiate between game-states, and for instance pawn-structure properties and king-placement. And between moves of a defender versus a attacker (or weaker versus stronger side) on one wing or center, and how does it interact with all apects of the evaluation.


In Monte-Carlo tree search, there is no position-evaluation function. The only heuristics are in move evaluation for ordering/pruning. Each branch of the tree is expanded until a terminal position is found. It is scored as a win or a loss. That is what makes MC search so special and cool. A lot of complex notions, such as King safety, should automatically "emerge" from the search. I wonder what kind of playing style it would produce in chess. Certainly not materialistic.

Some notions automatically emerge from the search when doing alpha-beta too, but I feel that what MC does is much more powerful. In a classical chess program, you have to compensate the short-sightedness of the search with evaluation features that anticipate what will happen in the long term. With MC search, the search goes to the end of the game, so the search is not blind to long-term considerations.

Rémi
Back to top
View user's profile Send private message Visit poster's website
Display posts from previous:   
Subject Author Date/Time
A Go algorithm for chess programmers to try ? Rémi Coulom Sat Jun 02, 2007 8:32 pm
      Re: A Go algorithm for chess programmers to try ? RVisitor Sat Jun 02, 2007 8:44 pm
      Re: A Go algorithm for chess programmers to try ? Daniel Mehrmann Sat Jun 02, 2007 8:57 pm
      Re: A Go algorithm for chess programmers to try ? Vincent Diepeveen Sun Jun 03, 2007 2:05 pm
            Re: A Go algorithm for chess programmers to try ? Rémi Coulom Sun Jun 03, 2007 2:18 pm
                  Re: A Go algorithm for chess programmers to try ? Vincent Diepeveen Sun Jun 03, 2007 2:48 pm
                  Re: A Go algorithm for chess programmers to try ? Dann Corbit Tue Jun 05, 2007 6:58 pm
                        Re: A Go algorithm for chess programmers to try ? Dan Andersson Wed Jun 06, 2007 7:21 am
                              Re: A Go algorithm for chess programmers to try ? Rémi Coulom Wed Jun 06, 2007 7:37 am
                                    Re: A Go algorithm for chess programmers to try ? Dan Andersson Wed Jun 06, 2007 7:58 am
      Re: A Go algorithm for chess programmers to try ? Alessandro Scotti Tue Jun 05, 2007 5:54 pm
      Re: A Go algorithm for chess programmers to try ? Gerd Isenberg Wed Jun 06, 2007 1:50 pm
            Re: A Go algorithm for chess programmers to try ? Rémi Coulom Wed Jun 06, 2007 2:40 pm
                  Re: A Go algorithm for chess programmers to try ? Gerd Isenberg Wed Jun 06, 2007 6:05 pm
                        Re: A Go algorithm for chess programmers to try ? Rémi Coulom Wed Jun 06, 2007 6:21 pm
Post new topic    TalkChess.com Forum Index -> Computer Chess Club: Programming and Technical Discussions

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum




Powered by phpBB © 2001, 2005 phpBB Group
Enhanced with Moby Threads