Author: Dann Corbit
Date: 13:02:52 06/04/02
Go up one level in this thread
On June 04, 2002 at 15:54:35, Robert Henry Durrett wrote: [snip] >After thinking about what I wrote above, it occurred to me that my questions may >appear somewhat dumb. > >I assume that chess computers, like humans, evaluate positions after each move >in a search, in order to obtain the next moves in the tree. How else would they >get the next moves? > >But your bulletin and others seem to be saying that the position evaluation for >the initial position [after a move but before any search] is evaluated much more >completely. Am I reading this right? If the root node is analyzed to 10 plies depth, it's immediate children are analyzed to nine plies, the children of the children are analyzed to 8 plies, etc. down to the leaf nodes. When we get to the leaf nodes, we just look for a quiet spot (no more obvious captures, checks or other nasty things that go bump in the night) and return that score. So the farther forward in the tree we go, the less deeply the positions are analyzed. >If so, then maybe the extent of evaluation for later positions occurring DURING >the search should be different for different positions, at least sometimes. > >For example, in human analysis, there sometimes comes a position which is >perceived as "critical" in some sense by the human evaluator. In that case, the >human does "a deep think" on that particular position, or at least does a better >job at evaluating that position. I wonder if similar situations occur during >the search, at least for some engines. > >In essence, the question is whether or not the amount of effort expended in >evaluating positions might be deliberately different for different positions. > >I apologize in advance if I am not making myself clear. > >Bob D.
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.