Author: Mogens Larsen
Date: 05:20:34 08/27/02
Go up one level in this thread
On August 27, 2002 at 07:53:58, Arturo Ochoa wrote: >I have a single question to this interesting analysis: Why three different >engines and now Yace are establising a remarkable difference in this kind of >Matches? > >I would like to know: The Match Yace is not 12-6 for Yace with Book against Yace >without Book. > >Is this not remarkable? In other engines matches of the Top, the difference was >12-8, 13-7. Is this not remarkable? I'm not quite sure what you mean. Are you saying that Yace does better with or without book in your tests? I've done a limited experiment and Aaron did a more extensive one. Neither with advantage for the one using a book. Of course this might say something about my book :-). >The problem is real: The premise established by him was: A engine can solve the >majority of the opening problems and an engine with book is not better that the >same engine without book. I disagree with both premises. The reason I find identical engine testing dubious is the similarity of the games. Even with position learning, the engine without book will suffer. Mainly because it'll diverge too late. That's why I prefer the comparison "autogenerated/amateur book vs. commercial book". A matter of taste, I know. The gauntlet approach, eg. 2-4 games against 20-40 engines, will avoid some similarity and the deficits of learning. And simulate a tournament. >Ok, let´s do it by this way. If it is waste of time to show that real statistics >are not valuable, then I won´t continue matches of this kind. I think you should continue. After all, my criticism might be wrong. Just wanted to emphasize that there are different methods. >Post Data: My compliments for Dieter, Yace has impressed me. :)) I agree. Too bad the SSDF doesn't feel the same way :-(. Regards, Mogens
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.