Author: Rémi Coulom
Date: 07:58:45 04/11/03
Go up one level in this thread
On April 11, 2003 at 02:33:12, Dann Corbit wrote: >On April 11, 2003 at 02:18:43, Albert Bertilsson wrote: > >>Hi! >> >>When testing my new engine against my old engine I'd like to have some >>estimation of how probable it is that the engine has become stronger. >> >>How do I calculate that? I guess there is some formula, I'd like to learn it. >> >>For example new version beats old version by 26-24 in 50 games. Is it stronger >>(with some probability I guess)? If possible, how much stronger? How many games >>should be played to get fairly accurate numbers when the engine is only a little >>better? >> >>I guess 10 games is to little but I can't do any development if I need 500 games >>to know if it has improved. > >http://www.stevemaughan.com/whoisbetter.htm > >Also elostat There seems to be a bug in Steve's program. For instance, 49% with 10 losses requires 11 wins according to this program. This is obviously wrong (10 are enough). It looks like his program returns the right answer + 1. http://remi.coulom.free.fr/WhoIsBest.zip Provides theoretical explanations and is probably a little better ;-) Rémi
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.