Exact math · both wheels · no sampling

How many spins until every roulette number comes up?

On average 155.5 spins on a European wheel (160.7 on American) before every number has come up at least once. In any 37 spins, about 13.4 numbers don't appear at all.

155.5
Spins to see all 37
average, European
147 spins
Even odds by
all 37 seen
13.4
Unseen in 37 spins
numbers, on average
8 spins
First repeat
more likely than not

How long until every number comes up

This is the coupon collector’s problem. With s numbers already seen, the next new one takes 37/(37 − s) spins on average. Add those up from 0 to 36 and you get 37 × (1 + 1/2 + … + 1/37) = 155.46 spins. On an American wheel it is 160.66. The last few numbers do most of the damage: the final one alone takes 37 spins on average.

Coverage of the European wheel after n spins
SpinsDistinct numbers (avg)Unseen (avg)All 37 seen
3723.613.4<0.1%
7432.14.90.2%
10034.62.46.4%
15036.40.653.1%
20036.80.285.5%
30037.00.099.0%

Half of all sessions see every number by spin 147, and nine in ten by spin 215. Probabilities in the last column come from an exact Markov chain over the count of numbers seen.

Sleeping numbers

A “sleeper” is a number that hasn’t come up for a while. The chance a particular number misses n spins in a row is (36/37)n. In any window of 37 spins, more than a third of the wheel sleeps.

Probability a particular number does not appear in n spins
Asleep forEuropean1 inAmerican1 in
37 spins36.29%2.837.28%2.7
50 spins25.41%3.926.36%3.8
74 spins13.17%7.613.90%7.2
100 spins6.46%156.95%14
150 spins1.64%611.83%55
200 spins0.42%2400.48%207
300 spins0.03%3,7130.03%2,982

An overdue number is not due

A number that has slept 200 spins comes up next with probability 1/37, the same as one that hit on the last spin. The wheel doesn’t keep score. The same holds in reverse for “hot” numbers: on a fair wheel a hot run is variance, not signal. Both ideas are tested as betting systems, on the hot-number method and betting against a streak. Telling real bias from noise takes thousands of logged spins; the wheel-bias sample-size study works out how many.

Repeats come early

The flip side, and the birthday problem in disguise: some number repeats far sooner than intuition says. The chance that the first k spins are all different is 37/37 × 36/37 × … × (38 − k)/37.

Probability at least one number repeats within the first k spins
SpinsEuropeanAmerican
38.0%7.8%
415.4%15.0%
524.6%24.0%
634.8%34.0%
745.3%44.4%
855.7%54.6%
965.3%64.2%
1073.7%72.7%
1286.5%85.7%
Check it against the wheel.The hot/cold heat map on the betting layout shades every pocket by recent hit frequency. Run a few hundred spins and watch sleepers wake up on schedule.