Author: José Carlos
Date: 10:50:59 01/11/02
Go up one level in this thread
On January 11, 2002 at 13:16:37, Dan Andersson wrote: >Some hefty ply depths you are mentioning. I happen to agree in principle with >them. But do you really think a program would survive if it searched eight ply >less? Just joking:) Not at all, of course, but what I mean is that a _big_ difference in eval quality _might_ compensate a big difference in search. > The problem is that even the best evaluation available today >is 'always' wrong, it will asses the static features but will have no clue of >the future of the position. This is true for my eval, but I can't speak for others. I suspect there are rather good evaluations out there in programs like Tiger, Fritz or DIEP. But even my weak program has some eval terms (most programs do) that 'anticipates the future', such as some geometric rules to detect if a king can stop a pawn or if a king will suffer a strong attack. Also, a good eval can eventually help search in the sense that if you trust your eval you can do some prunning of clearly stupid moves, which you have to prove stupid by tactics otherwise. >Thus missing the mark more often than not. The real >mystery here is that current chess programs manage so well in the first place. >Well, it's really no mystery but surprising none the less. > >MvH Dan Andersson I think there are two ways of playing perfect chess: searching the whole game tree and having perfect eval and searching 1 ply. Both are obviusly impossible to perform but they make me think trying to get as close as possible to both is equally important. José C.
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.