Author: Ed Schröder
Date: 23:05:25 07/21/99
Go up one level in this thread
>Posted by Harald Faber on July 21, 1999 at 08:18:33: >That leads me to the following question: How do you judge which are the best >settings? If I give to you 100 games 40/120 against the SSDF-TOP5 with a sound >overall result, would that be OK? Or do you have a special (secret) set of >test positions with which you verify the "best" settings? I will make a good research environment available. The main item will be the build-in autoplayer in Rebel. Just load a personality and autoplay 50 60/60 games against Rebel10.5's default settings. If the 50 games match ends in favor of your own created personality than you have an indication your created personality maybe is better. Next I will offer an option that you can pre-program opening lines yourself and Rebel's build-in autoplayer will follow these opening lines one by one. An example is included. The User-Book "50-POS.RBM" contains the 50 opening lines Jeroen Noomen once posted here to test chess programs. It's all automatic, you load the User-Book and off you go. No influences of the book, no learner effects just engine versus engine results. Next I offer auto232 to play against other chess programs. The disadvantage is that you need 2 Pc's. Another disadvantage of auto232 is that you need to turn off ALL learners, the learner of Rebel as well as the games MUST represent pure engine versus engine fights. Games who are influenced by learning are of no value to measure real progress of the chess engine. Complete details at release time. Ed Schroder PS, if you think the research environment can be improved then now is the time to mention it.
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.