On Thu, 30 Mar 2000, Kimes, Dean W. wrote:

> No. If it was only a 20% difference over the whole of the night, then some
> fights you would outdamage the other party significantly. that's how random
> numbers work. Fights are far too small in sample size individually for the
> whole's difference to hold true unless it is large.
>
> Example I did run through. 20 fights vs blue aviaks with a ranger of equal
> level. I out damaged him on one fight. Odds say that would mean I was 25%
> behind him assuming a sample size ration that was present, 20-1. It was
> close, he did 33% more damage than I did. Same level, same weapons.

20 fights is an awfully small sample to be drawing this type
of conclusion. I can see what you're doing (assume normal
PDFs, find number of incidents counter to expectation, using
that figure out the translation in the two PDFs to have that
probability of counter-expectations), but it's still just a
roundabout way of deriving the two damage histograms. Just
find the two histograms directly via parsing.

--
John H. Kim
kim@...