Author: Steve Maughan
Date: 09:07:33 01/29/99
Go up one level in this thread
I think Larry Kaufman first proposed that you could reduce the differencial by 75% to get the same outcome in Comp v Comp as Human v Human ie if 500 ELO point differencial is required to never loose a game in human ratings then 375 ELO points would be required in Comp v Comp. Interesting and probably true that Comp v Comp require a smaller ELO gap to acheive the same score ratio. I would also take this one step further and say that the trend of a program being launched, acheiving a huge ELO that gradually reduces over time is also a result of this phenomenon. ie a program is better than most and therefore wins a disproportionate ammount of games for its true improvement. This gives it a large than deserved ELO improvement over the other computers. As other programs catch up in strength the ELO of the first program starts to drop as it is no longer acheiving the high results. Just an idea! What does everyone else think? Regards Steve Maughan
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.