Author: blass uri
Date: 05:44:46 06/24/00
Go up one level in this thread
On June 24, 2000 at 07:58:44, David Dahlem wrote: > >On June 24, 2000 at 02:26:28, blass uri wrote: > >>On June 24, 2000 at 00:01:57, Chessfun wrote: >> >>>On June 23, 2000 at 20:09:03, Jorge Pichard wrote: >>> >>>>On June 23, 2000 at 11:06:26, Chessfun wrote: >>>> >>>>> >>>>>Told yer. >>>>> >>>>>Thanks >>>> >>>>In your first mini match of 30 games at G/30 they actually ended tied at the end >>>>of the 30 games. Now if you take fritz 6a vs Fritz 6a and play 30 games at >>>>whatever time control against itself the side the will win it will always be the >>>>side of the opening that has the advantage. therefore, the only reasonable way >>>>to test two personalities settings would be to let one settings play 15 games >>>>with white pieces and then switch personalities and color and repeat the same 15 >>>>openings for the remaining of the 15 games. This method allows both settings to >>>>play the same opening 15 times with both side of the board. >>>> >>>>Pichard. >>> >>> >>>Again without sufficient quantity of openings it proves nothing more than >>>just letting them play from book. Some of the settings may be more positional >>>while others more tactical, some defensive some offensive. While the 15 openings >>>you happen to play could suit one of those settings. The same argument is >>>naturally also applied to Nunn but to me if enough games are played from book >>>then you need not get into worrying about positional strengths or tactical >>>strengths. IMHO 15 positions is not enough within the same program to say >>>setting A is better than B. This is apparent with your CMQueen++ settings, >>>while it wins some matches it also losses some the same as the other settings >>>do. >> >>I agree that 30 games are not enough to be sure which personality is better even >>if you use the same opening but I believe that 30 games when after 15 games you >>play the same positions out of book with opposite colours can give you more >>significant result. >> >>I agree that it is possible that one personality is unlucky to get positions >>that it does not know how to play but at least you cannot say that one >>personality may be unlucky to get bad positions or positions that both sides do >>not know to play. >> >>This is different than the nunn match because in the nunn match the positions >>are not from the opening book of the program so one program can "complain" that >>it was not programmed to know how to play them when in the personality >>tournament the positions out of book are from the opening book of the program. >> >>Uri >I thought the reason for running personality tests was to test the personality >settings, not the opening books. In my opinion, eliminating the opening book >from the testing gives more consistent results. I think the target was to test the personality together with the chessmaster's book. Both sides use the same book so eliminating part of the luck of choosing an opening by using something similiar to the nunn match with the chessmaster book gives more consistent results. Uri
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.