Author: Bertil Eklund
Date: 12:07:57 11/26/02
Go up one level in this thread
On November 26, 2002 at 13:16:05, James T. Walker wrote: >Hello Tony, >Thanks for your comments. First if you have ever adjusted the rating list >upwards I have never seen it. I suspect it was very early in the testing stages >to get close to a percieved parity with human play. I also suspect that if all >the adjustments are combined the result is definitely down. I also suspect the >reason is because of the exageration of comp/comp games which keeps making the >top ratings higher that they would be if actually calibrated to human ratings. >That is all the point I was trying to make. You ARE trying to make the ratings >resemble human ratings and not JUST trying to compare computers to computers. >The error bars are probably accurate to the data they are fed with. I see no >reason for them to be wrong although I have not tried to check them. Would you >explain exactly what Thoralf has told all testers to do or not do? Can you tell >what exactly is the organization method of creating the data. >Jim Hi! The simple reason for the adjusting downwards are that almost every human are used to computer programs nowadays. In the past most of the human games came from tournaments, today most of the games are from matches when the human prepares day and night against it. In example when I play against Mach3, I performs maybee 200 elo better today vs when I bought it. The same goes for the big guys when they play Fritz and co. If Kramnik hadn't played months with Fritz and other programs I believe he should have lost. Thoralf organizes the meetings so we can have a good mix of opponents not in any scientific way but with common sense and fairness in mind. Bertil
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.