Published · · Last reviewed · · 6 min read

English · 한국어

Until every number has appeared: 31.5 draws on average

Lotto 6/45 draws six numbers at a time. How many draws does it take for every number from 1 to 45 to have appeared at least once? Many people estimate "45 ÷ 6 = 7.5 draws". The exact average is 31.5 draws, four times as many. The record agrees: starting from round 1, all 45 numbers had appeared by round 27, and when the 1,247 rounds are cut into blocks that restart each time all 45 have appeared, the 39 blocks average 31.9 draws. This page shows where those numbers come from and what a number that "has not appeared for a long time" looks like.

Why 31.5 and not 7.5

7.5 is the average wait for one particular number to appear (45 ÷ 6). For all 45 numbers to appear you have to wait for the slowest one. New numbers arrive easily at first; near the end, finding the few missing ones takes a long time.

Counting it out: the chance that a given number is not in a draw is 39/45 = 13/15 (about 86.7%). The chance that it stays out for k draws in a row is (13/15)k, so after k draws the expected count of numbers that have not yet appeared is 45 × (13/15)k.

Expected number of numbers that have not appeared after k draws
Draws kA given number still missingExpected missing numbers (of 45)
186.7%39.0
548.9%22.0
1023.9%10.8
205.7%2.6
301.4%0.61
400.33%0.15
500.08%0.04

After 10 draws about 34 of the 45 numbers have appeared, but 10.8 have still never come up. After 20 draws, 2.6 are still missing. The slow arrival of the last numbers is what lifts the average from 7.5 to 31.5.

The last numbers use up the time

Given that some numbers have already appeared, the expected number of further draws until all have appeared can be computed with a Markov chain whose state is the count of numbers seen so far.

Expected further draws until all 45 have appeared, given s numbers already seen
SeenMissingExpected further draws
04531.50
6 (after 1 draw)3930.50
242126.20
36920.42
40516.56
43211.02
4417.50

With one number left, the average wait for it is 7.5 draws (45 ÷ 6). With five left it is 16.6 more draws, more than half of the whole expected time (52.6%). By contrast the first draw fills 6 numbers (13% of the set) and uses 1 draw (3% of the expected time). The easy part passes quickly; the time piles up in the hard end.

If numbers were drawn one at a time (the classical form of the coupon-collector problem) the expected number of picks would be 45 × (1 + 1/2 + … + 1/45) = 197.8. Lotto draws six distinct numbers, so 31.5 draws is 189 numbers, slightly fewer, because within one draw no number repeats.

How variable is that time?

An average of 31.5 draws hides a wide spread. In the exact distribution the median is 30 draws, the 10th percentile 22, the 90th percentile 43, the 99th percentile 59, and the standard deviation 8.7. The chance of finishing within 27 draws is 36.8%, so the first 27 rounds of the real history were unremarkable.

The real record was used like this: count from round 1; when all numbers have appeared, close the block and restart from the next round; repeat to round 1,247. This gave 39 complete blocks (the last 2 rounds unfinished) of lengths from 18 to 67 draws, with mean 31.9 and standard deviation 8.75, very close to the theoretical 31.5 and 8.72.

Draws until all 45 numbers have appeared: theory and the 39 real blocks
BasisMeanStandard deviationShortest blockLongest block
Theory (exact)31.508.72--
Real (rounds 1–1247, 39 blocks)31.928.751867

Run it yourself

The tool below repeats Lotto 6/45 draws until all 45 numbers have appeared, 2,000 times, and compares the mean and spread with the exact values.

This tool needs JavaScript.

One number on its own

Looking at one favourite number instead of all 45 changes the story. Its chance in a draw is 6/45 = 13.3%, so the average wait is 7.5 draws, but even that wait varies. Half of the time it appears within 5 draws, 90% within 17 draws and 99% within 33 draws. Put the other way, a number picked at random has a 1% chance of staying away for more than 33 draws, and among 45 numbers it is reasonable for about one to do so. A single number's wait can be short or long; the wait for the whole set is long because the latest number decides it.

Other lotteries

The same method gives the expected number of draws until every number of other pools has appeared. A larger pool and fewer numbers per draw both make it longer.

Expected draws until every number in the pool has appeared (exact)
Lottery (pool)Numbers per drawExpected draws
Lotto 6/45 (Korea)6 of 4531.5
Loto 6 (Japan)6 of 4329.7
Loto 7 (Japan)7 of 3720.8
EuroMillions main numbers5 of 5043.6
EuroMillions Lucky Stars2 of 1218.1
UK Lotto6 of 5944.3
SuperEnalotto6 of 9074.5
Six-sided die1 of 614.7
Both sides of a coin1 of 23.0

This table also uses only each game's number format. It ignores details such as how the Jolly number of SuperEnalotto would be counted. The die (14.7 rolls) and the coin (3 flips) are included to show that the same formula holds for small pools.

How long is "a long time without appearing"?

This calculation touches the old belief that a number missing for a while is "due". The chance that a given number stays out for 20 draws in a row is (13/15)20 = 5.7%. Across all 45 numbers it is normal that a few have been out for about 20 draws (2.6 in the table at k = 20). In rounds 1 to 1,247 the longest any number stayed out was 73 draws. Generating 300 random histories of the same size gives a median longest gap of 64 draws and a 5–95% range of 55 to 82, so 73 is inside the range (the same record is treated separately in Hot and cold numbers).

This does not make a long-absent number more likely next time. Each round is independent of the last, so after 73 rounds off or after one, the chance that the number appears in the next round is 6/45. This page tells you that all numbers take 31.5 draws on average to appear; it tells you nothing about the next draw. How the feeling of waiting for a number turns into the gambler's fallacy is covered in The gambler's fallacy.

Where else the same calculation applies

This is the coupon-collector problem, known for a long time and met in everyday life. How many pulls to collect every item in a gacha (Loot box probability), how many rolls until every face of a die has shown (14.7 for a six-sided die): all the same structure. The shared lesson is that the last one or two items take a large part of the total time. When the final item in a collection stubbornly refuses to turn up, that is not bad luck; it is mathematically scheduled.

Precision of the real-history comparison

The theoretical values are exact (a probability process whose state is the count of numbers seen); the simulation only double-checks them. The 31.9 average from the real record rests on only 39 blocks, so its standard error is about 1.4 draws (8.75 ÷ √39), and a gap of 0.4 from the theory is within chance. The data are the main numbers of rounds 1 to 1,247 published by Donghaeng Lottery; the bonus number is not counted. Only aggregates are published.

LottoFlick is independent and not an official lottery operator. Numbers are random and cannot be predicted.

Sources and verification