Author: Uri Blass
Date: 05:56:13 12/11/02
Go up one level in this thread
On December 11, 2002 at 08:35:10, Nicolas GUIBERT wrote: > >> >>I do not think that it is easy to do it and I do not say that evaluation is >>unimportant. >> >>My point is that the quality of the evaluation is more important than the >>quantity. >> >>I believe that the quality of the evaluation of movei is better than part of the >>amateurs with bigger evaluation thanks to better choice of weights and better >>decision of what to evaluate. >> >>Uri > >Yes indeed, a well optimized evaluation function can do much better than a heavy >one with many useless features and bad weights. > >Interactions between the different parameters is also something important. > > >Another related thing : > >You can do 99% of the evaluation job with 2000 lines. > >When you want 99.9%, you need 20 000 lines. > >And then it is far from being over. At least for draughts :o) But I assume it is >the same for most games. I agree and the point is that I believe that the difference between 99.9% and 99% is smaller than the difference from searching one ply deeper so if you believe that you can make your program search 1 ply deeper than it is better to work on it. The difference from searching one ply deeper is still significant in chess. Uri
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.