Author: Dann Corbit
Date: 20:18:28 11/13/01
Go up one level in this thread
On November 13, 2001 at 23:08:14, Robert Randolph wrote: >I am currently working on writing a new evaluation for my program.. I have been >contemplating an eval based on positional aspects and no direct material count. > >For instance, a white knight on H1 would be worth maybe .2, but a white knight >on E5 would be worth 3.2 Of course these are my actual eval values, but examples >to better iillustrate my idea. > >Has this been done before with any success, or lack there of? > >If it was a plausible idea (as i beleive it could be) what would your >suggestions be for positional advantages and disadvantages in this system, as >they would be fairly differing than those of evals that also rely on material >count. [D]7k/b7/4N3/8/7q/8/5P2/5KRN w - - I think it is an interesting idea, but I think it has holes too. A piece either has "potential energy" or "kinetic energy" in a sense. When we move a piece to a strong, protected outpost, it has a sort of "active" or kinetic energy. When it's sitting like a lump in some stupid location, it has potential energy, in case we should move it some place useful. Just using the square numbers won't work (see the above diagram). I think that the actual value of a piece as far as it's "static worth" will be a function of the depth in plies of analysis at hand. If you analyzed 50 plies forward, you would not need to store *any* static value for the piece -- after all, you would see all the good things it could do in the future. On the other hand, if you only look ahead 2 plies, you *must* have the potential energy for the piece accounted for somewhere or you will throw it away without cause. Of course, I could be completely wrong. It's just a gedankenexperiment right now.
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.