Author: Fernando Villegas
Date: 13:55:22 01/02/99
Go up one level in this thread
On January 02, 1999 at 11:07:56, Enrique Irazoqui wrote: >On January 02, 1999 at 10:18:29, Mike Castañuela wrote: > >>Very interesting observation, in fact i think that since far years of G3 >>the engine "pure" improvement in general has been asyntothycal (almost). >>You said it first! (if at the end of tournament the established thing is >>fulfilled) > Hi Mike and Enrique: How could be other way? At the current level programs already have, to improve not asyntothically but with real jumps would means for the program to become the champ of the universe in a couple of years or less. Else: part of the discussion has its root in a semabtical issue. If you consider hash table usage, better books, etc as part of the program, improvements are important nevertheless, but if you pick up just the source code as the real thing, then maybe we see few changes. Anyway, the very nature of changes needed to improve from, let us say, 2550 to 2580 or 2600 level, are of such a "finesse" that I see very diffcult to grasp them just with the usual test and even with many games between computers. And surely very difficult also for the programmers that are trying to get them. In this matter the only enough good test would be an impossible one: a very long match between the programs againts real GM at 40/2 hours rythm. fernando >It has been discussed sometimes and I hope these games can bring some clarity. I >remember that 2 years ago Bob and I ended mad at each other after arguing about >whether Rebel's jump of 50 Elo points in the SSDF list in only one year was real >(me) or not (Bob). Have I ever been wrong? :) > >Enrique
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.