Author: Jorge Pichard
Date: 03:40:52 09/13/04
Go up one level in this thread
On September 13, 2004 at 06:23:25, George Tsavdaris wrote: >On September 13, 2004 at 02:29:47, Jorge Pichard wrote: > >> >>Jonny 2.70(Different Arena Opening Books) versus Movei00_8_251s >>------------------------------------------------------------------------------- >> >> a. (Wizard -v02) 18.0 to 12.0 = 30 games >> >> b. (Mainbook of Harry) 16.5 to 13.5 = 30 games >> >> c. ( A Runge-v01) 10.0 to 10.0 still running to equal 30 games >>----------------------------------------------------------------------------- The lower bottom two Openings don't have a chance to win 10 games in a row. >> d. (O-deville.pgn) 8.0 to 14.0 in favour of Movei00_8_251s >> >> e. (Trappole_v29.abk) 8.0 to 12.0 in favour of Movei00_8_251s >> (Gabriel LEPERLIER Book) >> >> I am now continuing with A Runge-v01 book for another 10 games. >> > > I suggest to play with each book 40 games to start having a better indication >of what is happening. Or anyway play at least 30 games each book, so you just >have to play with A Runge-v01, O-deville, Trappole_v29 another 10 games. Then >your test would start having some value. And by increasing the number of games >would be statistically more and more significant. Please take a look at the first two they both have played 30 games, so far with the A Runge-v01 book another 5 games has been played and I will let it play another 5 games to total 30 games. The lower bottom two Openings don't have a chance to win 10 games in a row. PS: I will give you a hint, the Wizard-v02 has the best performace so far against Movei :-) Jorge > So at the end we would know what book is better. But is really this that we >would know? Or we would just know what book is better for Jonny 2.70 against >Movei00_8_251s. And who can guarantee that this book would be better for List >5.12 too? And what about time controls. If you change the time controls, then >what is happening. Would the selected as the best book, be the best again? > > Unfortunately we can't answer to these questions unless we would make giant >tests for every combination. Yet by accepting the fact that different engines >have almost the same behaviour and that they play the same at all time controls >we would gain a huge amount of time, and by a simple test as yours we would be >able to know some general things. But the 2 aforementioned facts are totally >wrong. So no one should say that the "winner-book" of your test is the best >since this is a (false) generalization.
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.