So... If you ask a mathematician, you don't need a huge sample size (assuming the distribution is homogenous). A few hundred upwards of a thousand results are enough to extrapolate useful results, and more sampling than that, regardless of the population size, are usually just used to try to...