This question is typical of people who don't really know stats. "I want the smallest number of draws to know with X% confidence" is just so...argh!
https://stats.stackexchange.com/questions/664572/experimentally-verifying-frequency-of-a-multinomial-distribution?noredirect=1#comment1252264_664572
https://stats.stackexchange.com/questions/664572/experimentally-verifying-frequency-of-a-multinomial-distribution?noredirect=1#comment1252264_664572
Comments
On the other, fuck it sample until you get a needed precision (or area of the ellipsoid) but then how do you think about area of an ellipsoid