Author: Jonathan Lee
Date: 18:01:37 02/28/00
Go up one level in this thread
On February 28, 2000 at 18:55:57, chris sergel wrote: >I have seen different ratings of computer programs and seen >tournaments wherein different people have different results >in judging computer strength. >Would it be useful to not simply give a rating to computer programs but >to give it a rating with a variable? >For example, >suppose program x has a rating of 2300 at game 5 minutes, 2350 at game 30, >2425 at game 2 hours, and 2375 at game 5 hours. >I think that would be interesting information, but I haven't found it >anywhere. Perhaps it exists, but I'm not looking at the right place. >Or perhaps the variable might be CPU speed. If one program gained 100 points >by doubling processor speed at a given time, and another gained only 50 points >for the same faster processor, that would also be interesting to know. >I think, perhaps, the information has all been gathered but not displayed. I may give you some intuition on the rating versus time graph. Do you know what the graph looks like for "Y=(1/X)"? The blitz rating is usually "sky high". The computer "thinks too much" in longer time controls, thus the rating is lower than its blitz rating. The variables to look at is: 1. speed of the computer (obviously) 2. time controls 3. selective depth A high selective depth works great if human and computer have game in 5 minutes. I have Mephisto Genius 2 by Richard Lang and it is seven years old. Genius 3 is the benchmark for computer chess in my opinion. Jonathan (26th message)
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.