The question
Every lottery results site, this one included, publishes lists of "hot" numbers (drawn most often recently) and "cold" numbers (drawn least often, or missing for longest). People use them in opposite ways: some play the hot numbers because they seem to be on a run, others play the cold ones because they seem to be overdue. Both ideas make the same testable claim — that recent history changes what happens next. So we tested it.
How the test works
- Take a game's draws in date order, using only draws from its current format.
- Before each draw, count how often every main number came up in the previous 20, 50 or 100 draws — and only those. Nothing from the draw being tested, or anything after it, is used.
- Call the ten most frequent numbers "hot" and the ten least frequent "cold" (ties are broken by number, so the result is fully reproducible).
- Count how many hot and how many cold numbers the draw actually produced, then move on to the next draw and repeat.
Any fixed set of ten numbers is expected to match 10 × k ÷ n of each draw's numbers — 0.72 for Powerball (5 of 69), 1.00 for EuroMillions (5 of 50), 0.67 for SuperEnalotto (6 of 90). If hot or cold numbers carried real information, their averages would pull away from that line.
Results
Over 6,910 tested draws with a 50-draw window (and similar numbers for the other two), 36 of the 36 hot and cold averages below are within two standard errors of chance. The largest deviation anywhere is 1.7 standard errors. With the 50-draw window, hot numbers did slightly better than chance in 3 of the 6 games and slightly worse in the rest: no consistent direction, which is what noise looks like.
| Game | Window | Draws tested | Hot 10 | Cold 10 | Chance |
|---|---|---|---|---|---|
| Powerball | 20 draws | 1,367 | 0.695 (−1.4 SE) | 0.725 (+0.0 SE) | 0.725 |
| Powerball | 50 draws | 1,337 | 0.742 (+0.8 SE) | 0.724 (−0.0 SE) | 0.725 |
| Powerball | 100 draws | 1,287 | 0.732 (+0.3 SE) | 0.732 (+0.3 SE) | 0.725 |
| Mega Millions | 20 draws | 914 | 0.688 (−1.0 SE) | 0.734 (+0.8 SE) | 0.714 |
| Mega Millions | 50 draws | 884 | 0.719 (+0.2 SE) | 0.712 (−0.1 SE) | 0.714 |
| Mega Millions | 100 draws | 834 | 0.680 (−1.3 SE) | 0.676 (−1.4 SE) | 0.714 |
| UK Lotto | 20 draws | 1,105 | 0.991 (−1.0 SE) | 1.025 (+0.3 SE) | 1.017 |
| UK Lotto | 50 draws | 1,075 | 1.004 (−0.5 SE) | 1.011 (−0.2 SE) | 1.017 |
| UK Lotto | 100 draws | 1,025 | 1.009 (−0.3 SE) | 0.995 (−0.8 SE) | 1.017 |
| EuroMillions | 20 draws | 1,105 | 0.973 (−1.1 SE) | 0.986 (−0.6 SE) | 1.000 |
| EuroMillions | 50 draws | 1,075 | 1.027 (+1.0 SE) | 0.977 (−0.9 SE) | 1.000 |
| EuroMillions | 100 draws | 1,025 | 1.020 (+0.7 SE) | 0.980 (−0.7 SE) | 1.000 |
| EuroJackpot | 20 draws | 780 | 0.971 (−1.0 SE) | 0.981 (−0.6 SE) | 1.000 |
| EuroJackpot | 50 draws | 750 | 0.981 (−0.6 SE) | 0.989 (−0.3 SE) | 1.000 |
| EuroJackpot | 100 draws | 700 | 0.989 (−0.4 SE) | 1.030 (+0.9 SE) | 1.000 |
| SuperEnalotto | 20 draws | 1,819 | 0.637 (−1.7 SE) | 0.666 (−0.1 SE) | 0.667 |
| SuperEnalotto | 50 draws | 1,789 | 0.655 (−0.7 SE) | 0.677 (+0.6 SE) | 0.667 |
| SuperEnalotto | 100 draws | 1,739 | 0.652 (−0.8 SE) | 0.671 (+0.2 SE) | 0.667 |
"SE" is the standard error of the average: how far it would typically wander from chance in a test this size if numbers were picked blindly. Values between −2 and +2 are ordinary.
Why hot and cold numbers don't work
A draw machine has no record of previous draws. Each draw starts with the full set of balls, so a ball that has come up six times in the last month has exactly the same chance as one that has not appeared since spring. Believing a cold number is "due" is the gambler's fallacy; believing a hot number is "on a run" is its mirror image, the hot-hand fallacy. Both come from expecting short sequences to look evenly balanced, when genuinely random sequences are full of streaks and droughts — as the gap analysis on every Cold Numbers page shows.
This is consistent with the randomness tests: if draws are random and independent, past counts cannot carry information about the next draw, and the backtest confirms that in practice for all six games.
So what are hot and cold lists good for?
Describing the recent past, accurately — which is all they claim to do on this site. They answer questions such as "which numbers came up most this year?" or "how long since 7 was drawn?", and they make it easy to see how much ordinary variation a fair draw produces. They are not a way to choose numbers with better odds; every combination has the same chance. If you are going to play, the Responsible Play page is worth a minute.
Limits of this test
- It tests main numbers only; bonus pools are too small in most games to test with ten-number sets.
- It tests ten-number sets and three windows. Other definitions of "hot" are possible, but every one we could think of reduces to the same claim, and the randomness tests speak against all of them at once.
- UK Lotto uses Round 1 draws only, and each game only draws from its current format, so the number of tested draws varies by game.