Author: Edward Seid
Date: 06:22:34 06/17/01
(cross-posted on Rebel board) BACKGROUND I'm the operator of HiddenDragon(C) on USCL, which is Gambit Tiger 2.0. I've read with interest a recent thread on CCC concerning the proper standard to use in determining the strength of a computer chess program. One side states that an assessment of a program's strength should be based on its results against strong human opposition. The other side argues that since there is little existing data on man-machine slow matches, that a reliable conclusion cannot be made. PURPOSE As operator of HiddenDragon(C), I aim to collect game data and results of the Gambit Tiger 2.0 engine against FIDE-titled opposition using a slow time control for the purpose of accurately determining the true strength of Gambit Tiger 2.0 against humans under tournament-like conditions. METHOD I propose a series of mini-matches to be held against selected FIDE-titled human opponents on USCL. The first of these matches is scheduled to begin Thursday, June 21st against FilthyAnimal(FM). The time control will be 90 minutes + 60 second increment. (Note that this is slower than the the current FIDE control of 75 minutes / 40 moves + 15 minutes for the rest of the game) Currently, two games/week are planned. The first competitor to WIN 6 games will be declared the winnner. TOOLS The hardware used will be an AMD Athlon 1GHz processor, 768 MB RAM, Windows 2000 Pro. The program will be Gambit Tiger 2.0. All 3/4/5-piece tablebases will be used. SEEKING ADVICE I wish to optimize the settings for Gambit Tiger 2.0 and seek your advice. The currently proposed settings are: Permanent brain ON Transposition tables ON Use endgame databases ON Ignore 50 moves rule OFF Hashtable 192 MB Draw score -15 Anti-human play ON Engine priority NORMAL Style NORMAL Endgame cache size 16 MB Endgame depth 4 GUI books OFF Engine books ON will use gambit.tbk (renamed to ct.tbk) I appreciate all comments and suggestions. If you disagree with the above settings, please provide an explanation along with suggestions. I realize that this experiment is imperfect. In any case, I hope to add to the information available regarding man-machine matches. Ed Seid Operator of HiddenDragon(C) on USCL
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.