Time to do some statistics (*dusts off poor brain, opens wikipedia*)
I think this is a Bernoulli distribution so using p(1-q) for standard devation
Null Hypothesis: p(soul) = 0.02 (mean = 0.02, variance = 0.0196, standard deviation = 0.14)
Alternative hypothesis: p(soul) < 0.02
Data: P(soul) = 0.013666666 (N = 6000, successes = 82, sample mean = 0.0136666, sample standard deviation = 0.12802, standard error = 0.00176436)
(SE = sample standard deviation/sqrt(N) )
I think the appropriate test to do is a z test:
z = (0.2 - 0.0136666666)/0.00176436 = 3.5895925
We look this up on
http://business.statistics.sweb.cz/normal01.jpg and find that this result has about a 0.02% chance of occuring if the null hypothesis is true.
So we can safely reject the null hypothesis - with 99.98% confidence the soul drop rate is less than 2%, even with full luck and soul drop gettups.