Author: Dann Corbit
Date: 13:35:38 11/14/01
Go up one level in this thread
On November 14, 2001 at 16:07:41, Robert Randolph wrote: >As i was doing some thinking and i beleive that no material eval allows for a >much more coarse evaluation (in the majority of positions) which enhances both >mtd and PVS... But in order for this work effectively there would have to be a >dynamic understanding of the position as previously mentioned. Possibly the >addition of a totally sperate quiescent eval... havent worked this idea out yet >fully. > >Also the idea of peices with potential energy is no problem since with a coarse >eval, there is a posibility of a deeper ply (more speed in the search.. fewer >re-searches) and the peieces "potential" should be found within the deeper >ply... hrm. More input appreciated :-) I think losing the value for the pieces is a mistake. I have a plan to add piece weights as a function of depth searched (though I have not implemented it yet). It would require incredible depth to reliably estimate the value of pieces all the time. I suspect that they are going to have to retain *some* weight regardless of the search depth. Similarly for things like positional weight tables for piece locations. If we can truly see far ahead, these tables are not needed. But at what depth do they lose value? Is it sudden or gradual? Mostly, I have questions and not answers. In any event, I think losing all the piece weight will be a mistake.
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.