Author: Uri Blass
Date: 00:22:14 03/29/02
Go up one level in this thread
On March 28, 2002 at 17:34:17, Gian-Carlo Pascutto wrote: >On March 28, 2002 at 16:35:09, Uri Blass wrote: > >>I understand that it is not truth but you do not need to run >>1000 games in order to check if the computer is 3-10% slower >>that is translated to a loss of 3-10 points. >> >>If you do a change in the evaluation function then you may need >>1000 games in order to know if the program is better or worse but >>if you do a change that only change the speed of the program and not the >>evaluation or the search rules then >>the best way to test is not to play games. >> >>You can simply give the program to solve a test suite >>in order to check if it needs more or less time to do it. > >You do not need to play 1000 games to see that your computer is slower. > >You do need (approximately) such a great amount of games to positively >know the difference in speed generates an elo difference of 3-10 points. > >-- >GCP Being 100% faster means difference of 30-100 elo for all chess programs(except cases of bugs). It means that being 7% faster means differnce of 3-10 elo for all pchess rograms(I use the approximation 1.07^10=2) 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.