"Hot" and "cold" numbers are the most popular statistics in the lottery, and the most argued-about. So we ran the actual tests on every Powerball draw under the current number matrix: a formal test of randomness, and a backtest of whether betting on hot numbers would have worked. The answer is more interesting than either side of the argument usually admits.
Our sample is 1,393 draws from 7 October 2015 — when Powerball moved to its current 5-of-69 matrix — through 12 August 2026. That is 6,965 individual balls, so each of the 69 numbers should appear about 100.9 times if the game is fair.
First, a bug we had to fix
Our first run said Powerball was not random. The chi-square statistic came back at 108.5 against a 5% threshold of 88.3 — apparently strong evidence of biased balls. That result was wrong, and the reason is worth explaining.
Powerball only added Monday drawings in August 2021. Our archive, however, contained a Monday row for every earlier week too, each one silently copying the previous Saturday's numbers. Those 520 phantom draws counted real balls twice and manufactured a bias that was never in the lottery — it was in our database. (Lotto America had the same problem: 243 phantom Mondays before it began Monday draws in July 2022.) We removed them, verified that every deleted row was a byte-for-byte duplicate so no real draw was lost, and re-ran everything. As a check on the cleaned data, 1,393 draws produced 1,393 distinct number sets — zero exact repeats, which is precisely what randomness predicts.
We mention this because it is the whole lesson of number-frequency analysis: the pattern you find is usually a property of your data, not the lottery.
The randomness test
On the corrected data, the chi-square statistic is 79.9 with 68 degrees of freedom. The 5% critical value is 88.3. Powerball passes — the distribution of numbers is statistically consistent with a fair, uniform draw.
| Measure | Result |
|---|---|
| Draws analysed | 1,393 (Oct 2015 – Aug 2026) |
| Expected appearances per number | 100.9 |
| Most drawn | 61 (124 times) |
| Least drawn | 13 (75 times) |
| Range randomness predicts (95%) | 81 to 121 |
| Chi-square | 79.9 (critical value 88.3 — passes) |
The headline "number 61 has been drawn 124 times and 13 only 75" sounds like a meaningful gap. It isn't. Across 69 numbers you should see a spread roughly that wide by chance; the most and least common numbers in any fair lottery will always look striking in isolation. That is what makes hot-and-cold lists so persuasive and so empty.
The backtest: would hot numbers have paid?
A uniformity test only asks whether numbers appear equally often overall. The stronger question is whether hotness persists — if a number ran hot in the past, does it keep running hot? That is the actual claim behind the strategy, and it is testable.
We split the draws in half, took the ten most-drawn numbers from the first period, and counted how they performed in the second. They appeared 554 times against an expectation of 505 — about 2.35 standard deviations high. Repeating this across twenty combinations of split point and group size, the hot numbers beat expectation in 19 of 20, averaging +1.32 standard deviations.
That looks compelling. So we tested it against the only fair benchmark: simulated lotteries we know are perfectly random. We generated 100 artificial universes of 1,393 draws each, ran the identical twenty-test battery on every one, and asked how often pure chance produces a result as strong as ours. The answer: 3 times in 100.
What that actually means
A roughly 3% probability is, technically, the sort of number that gets called "statistically significant." We are not going to claim it here, for three honest reasons.
- We chose the test after looking at the data. We tried several split points and group sizes and reported the pattern across them. That is exactly the process that manufactures 1-in-30 results from noise, and it is why the finding needs an independent replication we cannot give it.
- The formal test of fairness passed. The chi-square result says no number is drawn more than chance allows. A borderline persistence signal sitting alongside a passing uniformity test is far more consistent with ordinary variation than with defective equipment.
- Modern draw equipment is rotated and tested. Lotteries use multiple ball sets and machines, swap them between draws, and weigh and certify them independently — which is precisely the regime designed to prevent a persistent bias from surviving over a decade.
Our conclusion, stated plainly: there is no reliable evidence that Powerball numbers are anything other than random, and a weak, self-selected signal is not the exception.
Why it wouldn't help even if it were real
Here is the part that settles it practically. Suppose the effect is genuine and hot numbers really do run about 10% above expectation. It still would not make a ticket worth buying, because the lottery's edge is not a few percent — roughly 40 cents of every dollar never comes back as prizes at all. A marginal shift in which balls appear cannot close a gap that large. The expected value stays firmly negative.
And chasing hot numbers carries a cost that is measurable. Hot numbers are popular numbers — published, shared and played by many people at once. If you win with a popular combination, you are more likely to split the jackpot. That is a real, quantifiable reduction in what you would take home, traded for a benefit that probably does not exist.
The takeaway
Every draw is independent. No number is ever "due" — a ball that has not appeared in 60 draws has exactly the same chance tonight as one that appeared last week, because the machine has no memory of either. Number-frequency statistics are genuinely interesting as a record of what has happened, and we publish them for every game on that basis. They are not a forecast, and the moment they are sold as one, someone is being misled.
If you want a decision the numbers actually support: pick unpopular combinations. It will not improve your odds of winning by a single fraction of a percent — but it will reduce how many people you share with if you do. See quick pick vs self-pick and our shorter explainer on hot numbers, or explore the frequency data yourself on any game's page.
Original analysis of NumbersIntel's own draw archive, covering Powerball's current 5/69 matrix. Statistical tests are described so they can be checked; figures will shift slightly as new draws are added. For information and entertainment only — not financial or gambling advice. The lottery is a negative-expected-value game; play for fun, never as an investment. You must be 18+ (21+ in some states). If gambling is a problem for you or someone you know, call 1-800-GAMBLER.