Author: Uri Blass
Date: 09:55:53 08/01/02
Go up one level in this thread
On August 01, 2002 at 12:52:23, Pointer wrote: >On August 01, 2002 at 10:32:58, Uri Blass wrote: > >>On August 01, 2002 at 10:05:48, Pointer wrote: >> >>>On July 31, 2002 at 14:24:46, Mike S. wrote: >>> >>>>On July 31, 2002 at 14:00:23, pavel wrote: >>>> >>>>>(...) >>>>>That rating list will suggest strength of the program based only on those 1000 >>>>>positions, and thus can be easily tweaked by anyone to play good only on those >>>>>1000 positions. >>>> >>>>By including 1.000 mini-books for those positions. Or use carefully tuned sets >>>>of parameters the engine activates, depending on the position. >>>> >>>>But not by normal means IMO. What is good for position #1, may be bad for >>>>position #2, etc.etc. "Easily tweaked by anyone" seems a bit optimistic :o) >>>> >>>>Referring to the idea itself, I think anything which is more complicated that >>>>"ready, steady, go!" won't ever be done by SSDF. Their methods will remain >>>>simple, common critizism and suggestions which I read regularly since I'm >>>>online, are usually ignored. >>>> >>>>Baseline: If you want a rating list done differently, you have to do it yourself >>>>(which many people do). >>>> >>>>Regards, >>>>M.Scheidl >>> >>> >>> >>>You both are wrong! >>> >>> >>>When you optimize your programm to play these 1000 different positions >>>successfully you generally increase the playing strength of your pogram. >>> >>>Even the 20 NUNN positions are enough for a generally increase of the playing >>>strength by tuning only these 20 positions. But the NUNN positions could >>>be a bit too special. You would get a NUNN-Elo instead of a average-chess-ELO >>>strength. >>> >>>Books are NOT allowed. >>>The programs play without any book, beginning at the selected starting >>>positions. (it's clear, i dont know how you got the idea of 1000 mini-books ...) >> >>You cannot forbid books because there is no practical way to >>know that book are not used without the source code. > > > > >from the theoretical point of view you are right! > >but in reality all people can observe the thinking line, and the chessprogrammer >has a VERY BIG job to make the thinking line consistence to the moves >saved in the 'secret' book. instead of wasting years of full-day work to >integrate 1000 tricky-secret-opening-books in his programm he should better work >at the chess-engine > > >> >>The only thing that you can do is to limit the >>size of the book together with the program >>because finding how much memory the program use >>is easy. >> >>I think that books are not very important and >>when you have many initial positions wasting time >>about building books is not going to give much > >maybe you are right, maybe you are not right. who knows .... > > >>unless the time control is blitz. > >why unless the time control is blitz ?!?!? because in that case a book may be more productive to save time on the clock and at longer time control a book may be counter productive because the engine may find better moves by itself. 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.