Author: Jorge Pichard
Date: 19:35:06 11/29/01
Go up one level in this thread
On November 29, 2001 at 21:13:48, Tina Long wrote: >On November 29, 2001 at 19:59:05, Jorge Pichard wrote: > >>In order for this match to reveal anything it needs to be continued up to game >>50. Since right now it is too close of a match. >> >>Pichard > >Statistically 50 is very little different from 39 games. > >You need to play 10,000 games in order..... > >Tina Long There has been several accasions where members from the SSDF extend the match to 50 games, whenever two programs have an equal score by the end of 40 games. But you are correct, that in order to have any sound evidence betwen two programs you need at least 500 games to notice a major difference. If after 500 games you still don't notice any major difference than they are equal. Another observation is, that even if two programs are perfect rivals the difference is always noticeable when they perform against other programs. That is why the SSDF is one of the most respectable rating system available. Pichard
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.