Talk:Construction/@comment-25275851-20141003172748/@comment-198.99.28.241-20141003205135

^ Let's say there are 2.2mil players and you want a statistically significant result:

95% confidence level with 3% margin of error you need a sample size of ~1000.

99% confidence level with 3% margin you need ~1800

There is no point going further simply because any more sample marginally improves the accuracy, and you need exponentially higher sample size to do so.