The original coupon collector’s problem poses the question:

If each cereal box contains a random coupon uniformly drawn from nn varieties, what is the probability that tt boxes must be bought in order to collect each coupon?

A subtly different question arises when considering the case that one does not know the total number of coupons. How many boxes must be bought before you know you have all of them?

Of course, you can never be absolutely certain. But say you buy one billion boxes, and get 333,495,391333{,}495{,}391 blue, 333,284,381333{,}284{,}381 red, and 333,220,228333{,}220{,}228 green coupons. I would be fairly confident that in fact there are three varieties of coupon! Here, you can test that intuition.

Samples drawn
0
Distinct coupons
0
Since last new

There is an extremely simple stopping rule which achieves a success rate of about 95%95\% uniformly over all nn. The messages after each round show you what I would have done in your situation using this rule. Can you determine the strategy?