Quote:
Originally Posted by BruceZ
It converges pretty rapidly to within 2%. It doesn't converge very rapidly to within 0.5% or to within 0.1%. The convergence for a given probability slows down with the square of how close you want it. For 0.5%, it takes 57600 for 3 standard deviations (0.27%). For 0.1%, it takes 1.44 million. The number n you need for a given number of standard deviations s and accuracy a when your chance of winning is p is
n = p*(1-p)*(s/a)^2.
I simply did the below. I don't see anything wrong. It's the chance to get at most 78% (or at least 82%) on an 80% event.
binomdist(2300,.78*2300,.80,true) = .9%
binomdist(4000,.78*4000,.80,true) = .09%
binomdist(5750,.78*5750,.80,true) = .009%
binomdist(7600,.78*7600,.80,true) = .0009%
Last edited by NewOldGuy; 08-02-2014 at 06:03 PM.