Data, records and curiosities
"Overdue" numbers are real gaps in a real draw history. Here is the distribution they come from, the expected longest drought, and why the wait after a long absence is unchanged.
Every lottery site has an "overdue numbers" table: the balls that have gone longest without appearing. The tables are accurate. The inference people draw from them is not, and the reason is a distribution most players have never seen written down.
For a game drawing k balls from N, a given ball appears with probability p = k / N in each draw, independently. The number of draws until it next appears follows a geometric distribution.
Two consequences do all the work:
1. Mean gap = 1 / p.
| Game | Matrix | p = k/N | Mean gap |
|---|---|---|---|
| UK Lotto | 6 from 59 | 0.10169 | 9.8 draws |
| Powerball white balls | 5 from 69 | 0.07246 | 13.8 draws |
| EuroMillions main balls | 5 from 50 | 0.10000 | 10.0 draws |
2. The tail is exponential: P(gap ≥ g) = qᵍ, where q = 1 − p.
For UK Lotto, q = 0.89831 and −ln q = 0.107246. So the probability a given ball is absent for the next 50 draws is:
q⁵⁰ = e^(−0.107246 × 50) = e^(−5.362) = 0.00468 — about 1 in 214.
That sounds rare. It is about to stop sounding rare.
The mistake is comparing an observed drought against the mean gap. You should compare it against the expected maximum gap, because the overdue table is by construction showing you the worst case out of thousands.
For M independent geometric gaps, the expected largest is approximately:
E[max gap] ≈ ln(M) / (−ln q)
Each draw produces 6 appearances, so the history contains about 1,000 × 6 = 6,000 gaps across the whole pool.
So in ten years of twice-weekly draws you should expect to see some ball somewhere go about 81 draws — roughly nine months — without appearing. Against a mean gap of 9.8 draws, that is a drought more than eight times the average, and it is the ordinary output of a fair machine.
Cross-check it directly with the tail formula. Expected number of gaps of at least 50 draws:
6,000 × q⁵⁰ = 6,000 × 0.00468 = 28
Twenty-eight droughts of 50+ draws in a thousand-draw history. And at 80 draws:
q⁸⁰ = e^(−8.580) = 0.000188, so 6,000 × 0.000188 = 1.1
About one drought of 80+ draws — matching the 81-draw estimate above. The two methods agree, which is the point of doing both.
q = 64/69 = 0.927536, −ln q = 0.075223. A thousand draws produce 5,000 appearances:
At three draws a week, 113 draws is about eight and a half months. Nothing is broken.
Here is the result that dismantles the entire overdue strategy.
P(gap ≥ a + b | gap ≥ a) = P(gap ≥ b)
The proof is one line: q^(a+b) / q^a = q^b.
In words: a ball that has been missing for 50 draws has exactly the same expected further wait as a ball drawn yesterday — 9.8 more draws in UK Lotto. The 50 draws of absence buy nothing. There is no accumulated pressure, no debt to be repaid, no mechanism by which a plastic sphere could know it has been idle.
This is the gambler's fallacy with the arithmetic attached, and memorylessness is precisely why the fallacy is a fallacy rather than merely an overstatement.
Two reasons, both worth recognising elsewhere.
The selection effect. You never see the balls having ordinary gaps. The overdue table shows the extreme of the pool by design, so the number at the top is drawn from a maximum distribution, not from the distribution of a typical gap. Judging it against the mean overstates its strangeness by a factor of eight.
The exponential tail is unintuitive. Human intuition treats "1 in 214" as effectively never. Multiply it by 6,000 opportunities and it becomes 28 occurrences. This is the same failure that makes near-miss patterns and clustered pairs look meaningful: rare events are guaranteed when the number of chances is large.
A drought becomes evidence when it exceeds what the formulas above predict for that game, over that many draws, across that pool size — and when it is backed by a physical mechanism. Ball-mass or machine-geometry bias is measurable directly, and operators test for it (ball weight and machine bias testing, is the lottery rigged?).
If you want the whole-pool version of the test rather than a single ball, the chi-square method is in testing hot numbers with chi-square.
Then take the longest gap on the page and put it through ln(M) / (−ln q). The overwhelming majority of "impossible" droughts land inside it.
And if a long-missing number is one you like: it is no more and no less likely than any other, but the numbers you choose do change who you split a jackpot with — see is 1-2-3-4-5-6 less likely?.
Last verified: 2026-08-29