Author: Martin Schubert
Date: 14:36:50 06/11/01
Go up one level in this thread
On June 11, 2001 at 14:13:15, Gian-Carlo Pascutto wrote: >On June 11, 2001 at 13:35:21, Aaron Tay wrote: > >>In this case, which is more accurate? Or is the latter based on a approximation >>of the normal distribution. > >As I understand it, the normal distribution is an approximation of >the trinominal distribution for a large number of games played. That's exactly right. So using the trinominal distribution is always more accurate then using the normal distribution. But both distributions doesn't have an solution for the problem that you don't know the probability of a draw. And all results depend on this probability. Martin > >>Aaron >>Someone who has only heard of biniomial and normal distribution.. > >Same here. > >The following may be of interest, it was in of the files the >accompanied the utility. > >>>>>>>>>>>>>>>>>>>>>> >I once wrote an article in the "Swedish ICCA" (now SSDF) magazine >PLY, about this topic. That article was completely based on the >normal distribution and later became the basis for how SSDF computes >the ELO's and the corresponding confidence levels. That was a long >time ago and modern methods are more or less based on computer power. >This time I had good help of my son David who opened my eyes for >computer simulations as a method to find the analytic description. >Maybe more practical and with some interesting possibilities. >...and fascinating! >>>>>>>>>>>>>>>>>>>>>> > >-- >GCP
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.