Author: Kurt Utzinger
Date: 04:38:52 12/16/01
Go up one level in this thread
Quite a lot of games are needed to say that program A is better than program B. Below an example of a 100-games-match between program A (=Gandalf 4.32g) and program X (rather strong program for which I am Beta tester) at time control 40'/40 under ChessBase-GUI, ponder=off, 32 MB hashtables, played in 2000 in exactly this range of order: games 1-10 3,0-7,0 [X wins] games 11-20 6,5-3,5 [Gandalf wins] games 21-30 5,0-5,0 [draw] games 31-40 3,5-6,5 [X wins] games 41-50 4,5-5,5 [X wins] games 51-60 3,0-7,0 [X wins] games 61-70 5,0-5,0 [draw] games 71-80 8,0-2,0 [Gandalf wins] games 81-90 7,0-3,0 [Gandalf wins] games 91-100 5,5-4,5 [Gandalf wins] Total 51 - 49 for Gandalf_432g Comparing this list you can put the question: what if the games 71-90 would have occured first in a match about 20 games only? Program X would have suffered from a disastrous loss [15-5], but at the end of 100 games, program X obtained a nearly even score. Regards Kurt
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.