Author: Jorge Pichard
Date: 04:52:12 02/23/01
Go up one level in this thread
On February 23, 2001 at 07:05:13, Uri Blass wrote: >On February 23, 2001 at 06:02:55, Jorge Pichard wrote: > >>On February 23, 2001 at 05:54:17, Jorge Pichard wrote: >> >>>On February 23, 2001 at 05:49:16, Jorge Pichard wrote: >>> >>>>With the best settings available for Nimzo 8 (Middlegame Mobility 55 and Style >>>>Allaround) instead of the default settings using my Athlon 800 MHz it seems to >>>>me that Nimzo 8 strength could be improved by as much as 30 points compared to >>>>its default settings. Dr.Chrilly you should take note of this and if is not too >>>>late, please modified Nimzo 8 settings for the remaining of the ICCP match. >>>I forgot to mention that Deep Fritz won the match, but it was not as bad as it >>>would have been with its default settings. >>I also understand that according to Christophe 50 games is not enough to proof >>anything, that I need at least 200 or more games, but I don't need 200 rounds to >>be beating by Mike Tyson, 10 rounds is sufficient for two professional boxers >>and in my case I might last 10 rounds if I train hard enough to run for 10 >>rounds. > >The number of games that you need is not constant and is dependent in the >result. > >I guess that everyone will agree in cases that original nimzo8 is losing 50-0 >against Deep Fritz when new Nimzo8 is winning 50-0 against Deep Fritz that 50 >games is more than enough to be convinced that the new setting is better. > >I do not think that you can claim that the new nimzo is better than the old >nimzo based on only one opponent and I think that you need to test both nimzo >against many programs because it is possible that one setting is 30 elo better >against Deep Fritz but 30 elo worse against another program. > >30 elo difference is a small number and 50 games are usually not enough to know >which version is 30 elo better. > >You can have a big probability to be wrong if you try to claim things like it >based on only 50 games(I guess that if the difference is 30 elo then you have >probability of about 30-40% to be wrong in saying which program is better based >on 50 games of both programs against Deep Fritz but I did not calculate it) > >Uri I also tested the modified Nimzo settings against Fritz6e or english version using my second AMD K6-2 500 MHz computer and the score at G\25 after 15 games was W6 L5 D4 I will leave it for another day until it reaches 30 games. But I can assure you that I am not the onlu one testing the modified Nimzo 8 settings there are two other persons at this time that are testing as well once we put all of our testesting scores together it will end up being at least more than 100 games total. Pichard.
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.