Author: Dann Corbit
Date: 09:30:55 06/03/02
Go up one level in this thread
On June 01, 2002 at 09:15:12, Omid David wrote: >The question was an algorithmic one: How to escape a full window search of >(-10000,10000) at any cost. In theory this shouldn't occur since after the >following steps: > >best = negascout(alpha, beta, depth); >/* re-search */ >if(best <= alpha) > best = AlphaBeta(-10000, beta, depth); >else if(best >= beta) > best = AlphaBeta(alpha, 10000, depth); > >The value should be between the alpha-beta bound. However in practice sometimes >it happens that after the re-search the result is still out of the bound. The >frequency of this depends on the evaluation function. While Genesis faces this >problem very rarely, other programs with simpler eval functions will more >frequently have to deal with the problem above. > >In none of the open-source programs I found an algorithmic solution to the >problem. Maybe this is the way it has to be! > >And, thanks Dann for all the code you posted, it saved me hours of navigation in >different programs' codes! I think pretty much everyone does the same thing. After all, the search failed. You have to find the right score.
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.