Author: Gian-Carlo Pascutto
Date: 13:19:34 04/26/02
Go up one level in this thread
On April 26, 2002 at 16:12:49, Ratko V Tomic wrote: >>I presume you mean +- 1/sqrt(N) ? >> >>>From that you can get the >>>upper and lower values for probabilities (normalized >>>scores). > >I was using +/-sqrt(N) for absolute error. The relative >(normalized to N) error is then 1/sqrt(N), as implied >in my next sentence. Ok, so if I get it right: for score 10/20 error is +- 4.47 95% confidence interval is 1.06 - 18.94 / 20 It looks to me like this will fall apart on the edges, i.e. the stat needs to take into account the scores are bounded into 0-20 interval. -- 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.