Author: Uri Blass
Date: 01:02:57 03/25/02
Go up one level in this thread
On March 25, 2002 at 01:11:21, Lyn Harper wrote: >On March 25, 2002 at 00:26:50, stuart taylor wrote: > >>Or 7.01 patch >>Because the 7.32 was only a calibration of 7.00 for 32 bits. And 7.01 was only >>an insignificant internet patch. >>I never even went further than 7.00 and nothing since was proven to be better. >>So really we have been waiting about 4 years. >>S.Taylor > As far as I'm concerned, chess programming in general seems to have > hit the wall about 4 years ago. Since then, only incremental increases > in playing strength. I disagree The improvement in software in the last 4 years are bigger than the improvement in software in the previous 4 years. I believe that chess programs hit the wall in 1994 when genius3 appeared The other programs were clearly weaker than genius3 at that time so there was practically almost no progress in the next 4 years. In the last 4 years there was a big progress in software and Fritz7 is more than 100 elo better than hiarcs7. My last purchase, CM8K, over a year ago, held it's > own against the best recently in a 100 round tourney.. won the tourney > in fact albeit by a small margin. Hiarcs7.32 was the worst program in that tournament and the result do not contradict the claim that it is 100 elo worse than Fritz7. > I admire the skills of a good chess programmer, but just get the > feeling we are relying on hardware improvements in recent times. The right comparison is between different versions of the same program and based on sarah's rating list there was a big improvement only in software in the last years. It is possible that the right personality of chessmaster8000 was clearly better than the top programs one year ago but nobody knew the right personality. Finding the right personality is part of the job of the programmer and if he failed to do it then it is unfair to claim that his old program was better at that time. I believe that other programs can be also improved by 50 elo if the programmer only change some numbers in the source code without changing a single line in the source code except some numbers that may influence the evaluation or the search rules. programmers may decide that every number of these numbers is a parameter of their program that can be edited by the users. Suppose that a user find a way to make the program 50 elo better 2 years after the program is released. Is it fair to call it an old program? I think that it is not fair because a work was done on the program. 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.