Author: Jorge Pichard
Date: 02:30:41 06/28/05
Go up one level in this thread
On June 28, 2005 at 05:07:28, Uri Blass wrote: >On June 28, 2005 at 02:42:21, Sune Fischer wrote: > >>On June 28, 2005 at 02:04:27, Reinhard Scharnagl wrote: >> >>>On June 27, 2005 at 21:58:51, Jorge Pichard wrote: >>> >>>>Just because a program is strong at standard chess doesn't mean that it will be >>>>strong at Chess960. For instance there is a noticeable difference between >>>>standard Pharaon 3.3 and Standard Freenze, whereas according to my test, >>>>FRC-Freenze is almost equal in strength to Chess960 Pharaon 3.3. >> >>I find that hard to believe, usually when things look too good to be true it's >>because you just haven't played enough games yet... ;) >> >>>This is not at all surprising to me. It reflects the unability of most engines >>>to UNDERSTAND positional implications during opening. This will be covered in >>>traditional chess by a) opening libraries, b) optimized weights in piece type >>>specific coordinate related bonus/malus tables. >>> >>>Smirf is avoiding the use of such tables to force the programmer (in that case: >>>me) to improve the program to a better positional understanding by AVOIDING >>>such intelligence murdering piece-coordinate 'optimized' tables, and of course >>>by switching from huge looking-up tables corrupted chess to Chess960's freedom. >> >>You tell 'em Reinhard! :) >> >>Design philosophy or not, the question is what is it worth in real play Elo. >>I think unfortunately not much. > >I agree. > >> >>Of course if the engine is used to having a good book and doesn't have any >>development terms on its own, then one could possibly be in for a grim awakening >>in FRC. > >I think that search dominates here and special developments terms for the >opening are not very important(I will be surprised if special knowledge for the >opening gives more than 100 elo in FRC and I expect it even to be not more than >50 elo). Are you implying that in standard chess Shredder 8 without an opening book can still beat Fritz 8 with its own opening book? Here is the Rating List of Fefruary, which indicates a difference of 50 rating points. Jorge > >Usually a program that plays the middle game well will also be good in Chess960 >because even if it gets inferior positions it can often get out of them and win >later. > >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.