Author: Dann Corbit
Date: 16:06:16 01/07/02
Go up one level in this thread
On January 07, 2002 at 18:39:44, Rajen Gupta wrote: >i think to be fair to all programmes every match on the ssdfshould be of roughly >40 games as otherwise it is easy to skew the ssdf results in favour of or >against a particular programme.i fail to see the logic of terminating the >testing so early unless it is a policy decision that will affect all matches >equally. It won't skew the results, except that it will make a wider error bar. You might think, "I'll play a single game against the top SSDF program and then if I win, I will stop right there." That would indeed give you a high ELO rating, but the error bar would be infinite. In order to have an acceptable error bar, you have to play a lot of games and this will level out the ELO rating. The system is self-correcting. You might have a good run of chance against opponent 'A', and also against 'B' but you won't keep having fortunate outcomes again and again and again unless you really are strong. Playing a large number of games against varied opponents brings the size of the error bar down quickly.
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.