Author: Robert Hyatt
Date: 21:13:13 10/27/02
Go up one level in this thread
On October 27, 2002 at 21:12:20, Ricardo Gibert wrote: >On October 27, 2002 at 20:58:25, Robert Hyatt wrote: > >>On October 27, 2002 at 18:55:25, Roy Eassa wrote: >> >>>On October 27, 2002 at 17:14:22, Ingo Lindam wrote: >>> >>>>Hello... >>>> >>>>I just thought about some questions from Bob that lead me to a problem. >>>>I know the engines should have as much knowledge as the programmer can give to >>>>it. But ofcourse in may searchtree may appear some positions I can't apply a lot >>>>(or in worst case any) of that knowledge: material is equal and all the other >>>>knowledge doesn't fit to current position (in the tree). Will my score 0 (or >>>>atleast very near to 0)? >>>> >>>>If yes...I just ask me whether this could lead the computer to optimize the best >>>>move sequence into complete knowing nothing about the position the sequence of >>>>moves end in... as more as I evaluate the positions for my opponent the same >>>>way? >>>> >>>>Shouldn't I better substract something from my score and add something on the >>>>opponents score in case I know nothing about the position? >>>> >>>>Ofcourse I know it is very probably that this is already done by every engine >>>>programmer and I just don't read anything about it yet or that I just make a >>>>mistake thinking about it now. >>>> >>>>Who can tell me? >>>> >>>>Ingo >>> >>> >>>I cannot answer your question, but may I attempt to reword it to see if I >>>understand: >>> >>>A program could determine the relevance of its knowledge to each position. >>>I.e., it could "know" that it does not understand some positions as well as >>>others. It could penalize the score of a position in the search tree by an >>>amount proportional to the lack of understanding. Ultimately it would then >>>penalize and thus avoid positions it understands the least. >>> >>>Is that about it? >> >> >>The danger is that the program is doing a very big search at today's speeds. >> >>If you leave a small "hole" in the positions your evaluation can produce >>scores for, then the search will head for that hole if the basic evaluation >>is bad, since those scores might be nearer zero... > > >These "holes" are unavoidable, since the positions the program does not >understand are being misevaluated anyway. Perhaps you may as well give such >positions some type of useful bias in score. As someone else pointed out, this >is what you already do by penalizing closed positions. It isn't the same thing, however. Avoiding blocked pawns is a definite issue to a computer. And saying "that is bad" will occasionally be wrong, until it is (one day) done correctly. But, because of the computer's unique strength in open position, this makes perfect sense as an evaluation term. But that is different than saying "I don't know what to do in this kind of position so I just will rely on the material score +/- some static offset. I don't have _any_ "intentional known holes" in my evaluation. And as I find unintentional holes, I try to fill them quickly, because a GM will fill them if I don't...
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.