Author: Tord Romstad
Date: 02:35:13 07/16/05
Go up one level in this thread
On July 16, 2005 at 00:26:56, Pallav Nawani wrote: >On July 15, 2005 at 18:29:28, Dann Corbit wrote: > >>The AEGT and CEGT contests assume the NUNN positions as openings, and so they >do not exercise the opening book of the program being tested. That is fine to >>measure engine strength, but it will not tell you about book+program and it >will not help you to prepare for that opponent (if it is a goal). > >Not all CEGT testers use Nunn positions. Some testers use a single book for all >engines. Actually, I think *most* CEGT testers use a single book for all engines. >>If we did a million games at G/1 second, we will have a lot of games but bad >>data. So longer games are (inuitively) better. > >The games will be low quality, but at such a time control nearly all programs >will have their performance degrade nearly equally, therefore, in general the >_ratings_ will still hold. Of course, there are some execptions here and there. At such fast time controls, I think there would be many exceptions. The rating list would look wildly different. I have one data point to offer: Glaurung 0.2.4 scores around 50% against Fruit 2.0 when testing at a time control of 4 1. This agrees rather well with the CEGT ratings, where Fruit 2 is currently 5 Elo points ahead of Glaurung. If I reduce the time control to 1 1, however, Glaurung's score drops to around 25%. I am not sure why; but probably I have some kind of time managment problem which is only relevant at extremely fast time controls. Tord
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.