Author: Jouni Uski
Date: 02:43:43 06/12/00
Go up one level in this thread
On June 12, 2000 at 05:30:50, pavel wrote: >Hi, > I know this question has been asked in this forum alot of time. But i didnt >followup the posts too well, thats why i am asking it over again. "How many >games between two programs will give and approximate elo differance?", now let >me clarify my position here. I am aware of these facts:- > >1)any day any program with a elo differance of 100 can beat each other depending >on the position right after the opening book. >2) some programs play well in shorter time controls while others play well in >longer time controls. >3) a set of (even 50) games is not strong enough to prove which program is >supirior. >4) last not least the question of validity of computer matches in single cpu >(whether its ponder=on or ponder=off!!). > >now facts apart. Is there still no way to find out lets say in a set of 100 or >150 or maybe 200 games between 2 programs in single cpu? > Note, that even 200 games isn't enough to find elo difference! You must play against minimum 10 different programs and prefeably 40 games in all matches. Some programs have so called Angstgegner and playing against them proves nothing (example Crafty17.7 and Junior6). Jouni
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.