Author: José Antônio Fabiano Mendes
Date: 05:31:20 02/24/00
Go up one level in this thread
On February 24, 2000 at 06:23:18, Martin Schubert wrote: >On February 24, 2000 at 05:16:48, Dann Corbit wrote: > >>ELO is self adjusting. >> >>Programs that learn will rise in ELO. >> >>Programs that repeat the same mistakes will drop in ELO. >> >>This is as it should be. Just like a human who learns. They are not static in >>ability either. Each game is a pinpoint in the sea of data. After each >>measurement, both pinpoints are disturbed. Neither sits where it was before. >> >>It could well be said that ELO tells us nothing about a contest between machines >>A and B of well-known ELO a-priori. >> >>It only tells us what a probable outcome should be. Not what will be. > >That's right. >But is had been said that ELO's of old programs one old machines are more >reliable because with these programes have been played more games. >This is an statistical argument. >After the Cadaques-Tournament you could read statistics of the tournament with >confindence intervalls etc. >My question: how can you calculate confindence intervalls without any >statistical conditions to hold? An interesting article about the so-called Bayes´ theorem: http://www.maa.org/devlin/devlin_2_00.html I enjoyed reading the article,maybe it can serve as a reminder of how trick statistics can be. JAFM
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.