The Coupon Collector's Problem
"Collecting the whole set takes far longer than the last few items, because the final ones almost never turn up."
"Collecting the whole set takes far longer than the last few items, because the final ones almost never turn up."
The coupon collector's problem asks how many random draws it takes to collect a complete set of distinct items when each draw yields one of the items uniformly at random. If there are n different coupons, the expected number of draws to collect them all is n times the nth harmonic number, which for large n is close to n times the natural logarithm of n. The striking feature is how the tail dominates: the early coupons arrive fast, but the last few are agonisingly slow, since once only one is missing each draw has just a one-in-n chance of being the one you need. It is the mathematics of why finishing a sticker album costs so much more than starting one.
W = \{ u \in L^p ([0, T]; X_0) | \dot{u} \in L^q ([0, T]; X_1) \}.Rated Empirical: grounded in studies or measurable evidence. The other 608.
The total wait is a sum of independent geometric waits, one for each new coupon. After you already hold some coupons, the chance that a draw brings a genuinely new one shrinks as the missing set shrinks: with k distinct coupons in hand, a new one appears with probability n minus k over n, so the expected wait for it is n over n minus k. Adding these waits from the first coupon to the last gives n times the harmonic sum, one plus a half plus a third and so on down to one over n. The last term, waiting for the final coupon, alone contributes an expected n draws, which is why the endgame is so long.
Erdős and Rényi showed the completion time, once centred and scaled, converges to a Gumbel distribution, so the collection time is concentrated around n log n with a heavy right tail.
If some coupons are rarer than others, the expected completion time only grows, since the rarest items come to dominate the wait, which the uniform formula does not capture.
The coupon collector's problem is a canonical result in probability with a long practical reach, into card and sticker collecting, software test coverage, ecological species estimation, cache and hashing analysis, and DNA sequencing coverage, wherever one must draw at random until every category is seen. Its central lesson, that completing the set costs vastly more than getting most of it and that the growth is n log n rather than linear, is a durable corrective to the intuition that the last piece should be no harder than the first. It also anchors the analysis of the last-item bottleneck in many randomised systems.
The standard result assumes each draw is independent and every coupon equally likely, an idealisation. When coupons have unequal probabilities, as promotions often arrange, completion takes strictly longer, because the rarest items dominate the wait, and the simple n-times-harmonic formula understates the true effort. The result is about the expected time and its distribution, which has a heavy right tail, so individual attempts vary widely around the average and a run can take far longer than n log n. It also assumes the set size n is fixed and known, and draws are with replacement from a stable pool.
It does not grow linearly with the number of items; the expected time is about n times the log of n, which is super-linear, so a bigger set is disproportionately harder. It does not say the last coupon is as easy as the first; the final one takes on average n draws by itself. And the clean formula assumes uniform, independent draws, so it is a lower bound on the effort when real-world items are unequally common.
Yes, as far as the evidence goes. The Coupon Collector's Problem is rated Empirical here: it rests on studies or measurements rather than on a saying.
Rated from two sources. 609 of the 1,116 entries carry this rating. How the scale works.
The best-known laws are the least likely to rest on measurement.
A classic problem of elementary probability with roots in the work of de Moivre and Laplace, given its modern distributional treatment by Paul Erdős and Alfréd Rényi in the 1960s.
8 languages, as recorded on Wikidata. These are names the idea already goes by, not translations we made. The whole index by language.
Everything above traces to en.wikipedia.org and brilliant.org: two secondary sources, accounts of the original rather than the original itself, which has not been located for this entry. Nothing is written from memory, and the method is written down. Spot an error or a better source? Suggest a fix.