"brumar_lv" <brumar_lv@...> wrote:
There may be no "perfect" definition, but this might work. The long
term is the number of games where a normal approximation applies ( to
Does anyone know what the formula is? I've looked, but can't find
it. Perhaps someone has a different definition? I know this is
Standard statistical tests for the equality of two probability distributions include the Kolmogorov-Smirnov test, Anderson-Darling test, and Wilk-Shapiro test. Here are Wikipedia links:
K-S: <http://en.wikipedia.org/wiki/Kolmogorov-Smirnov_test>
A-D: <http://en.wikipedia.org/wiki/Anderson-Darling_test>
W-S: <http://en.wikipedia.org/wiki/Shapiro-Wilk_test>
Here are links to the NIST's online Engineering Statistics Handbook:
K-S: <http://www.itl.nist.gov/div898/handbook/eda/section3/eda35g.htm>
A-D: <http://www.itl.nist.gov/div898/handbook/eda/section3/eda35e.htm>
W-S: <http://www.itl.nist.gov/div898/handbook/prc/section2/prc213.htm>
I might be able to produce some numbers for a few games with some more caffeine and motivation.
N0 gives you a definite number that I guess is useful, but it doesn't tell you how close you are to a normal distribution. Concluding there's an 84% probability that you will be in positive territory for a positive game assumes what brumar would like to prove. Still, that number probably isn't too far off for most games.
Mike
ยทยทยท
----------
Michael Peck
http://wildlife-pix.com