Why hot and cold lottery numbers do not work
Every lottery site publishes a list of hot and cold numbers, and the counts on it really do differ, sometimes by a lot. The thing almost none of them shows you is the figure that makes those differences interpretable.
The gap is real
Take Bonoloto. Across 3,538 draws in our archive, the most-drawn number has come up 89 more times than the least-drawn. Put that in a table, sort it, and the top looks meaningfully different from the bottom.
It is not a rounding error and nobody is making it up. The question is not whether the gap exists. The question is what size of gap you should expect when nothing at all is going on.
The number nobody shows you
Here is the trap. You are not looking at one number and asking whether it is unusual. You are looking at fifty-odd numbers and picking out the highest and the lowest. Take the extremes of any large group and they will be far apart, however identical the group is.
That expected gap can be calculated. For a pool of 49 numbers that all behave identically, chance alone puts the highest and lowest about 4.5 standard deviations apart.
| Game | Draws | Observed gap | Gap chance produces |
|---|---|---|---|
| Bonoloto | 3,538 | 4.6 sd | 4.5 sd |
| Mega-Sena | 3,086 | 6.7 sd | 4.6 sd |
| US Powerball | 1,434 | 4.9 sd | 4.7 sd |
| La Primitiva | 1,334 | 4.2 sd | 4.5 sd |
| El Gordo de la Primitiva | 554 | 6.0 sd | 4.5 sd |
Read that table across, not down. In every one of those games the observed gap sits close to the gap chance produces. The hot and cold lists are exactly as spread out as a fair draw makes them.
Two questions people mix up
How long has a number been missing? and what is the longest it has ever been missing? are different questions, and the second is routinely presented as the first.
A number can hold the record drought for a game and have been drawn last week. "Number 12 has gone 111 draws without appearing" reads as a statement about now. Usually it is a statement about 2019.
On our statistics pages both figures are shown separately and labelled, because the difference is the whole point.
What about overdue numbers?
"Overdue" is the same claim as cold, with a deadline attached: this number has not come up for ages, so it is owed one. Every site selling a hot-and-cold picker offers it. Here is what it looks like in the deepest archive we hold.
In 3,538 Bonoloto draws, the longest any number has gone missing is number 48, absent for 99 draws in a row. On the day that run reached 99, every overdue table in the world would have had it at the top.
It has been drawn 432 times in that window against an expectation of 433 — which makes it the 28th most drawn of the 49 numbers in the game. A number can hold the longest drought in the archive and be perfectly ordinary across the whole of it. Both things are true at once, and neither predicted the other.
Nor is a long gap unusual. 21 of the 49 numbers have gone 50 draws or more without appearing at some point. A drought is what this many draws of a game this size produces; it is the normal texture of the data, not a signal inside it.
The arithmetic underneath is the uncomfortable part. On the draw after a number has been missing 99 times, its chance is the same as it was on the first: the balls carry no record of the run, and nothing is being repaid.
The gambler's fallacy, precisely
A ball that has not come up for a year is not owed an appearance. The machine has no record of what it did last week, no mechanism for keeping score, and no way to act on one if it had. Each draw is independent, which means exactly this: knowing every previous result tells you nothing whatsoever about the next.
The reverse belief — that a hot number is "running" and will continue — is the same error wearing different clothes. Both assume the draw remembers.
Then why do we publish the tables?
Because they are a genuine record of what has happened, and because the alternative is leaving you to find the same list somewhere that attaches an implication to it.
We publish them with the comparison number attached, and we withhold them entirely when an archive is too short for the ordering to mean anything at all. On a young archive a ranked list is noise with a sort applied, which is worse than showing nothing.
Figures computed Saturday 24 October 2026 from our published archive.