Published · · Last reviewed · · 5 min read
English · 한국어
The birthday-number trap: same odds, different split risk
You may have heard that "many people play birthdays, so numbers up to 31 will be shared if they win". The claim mixes a correct part with an unverified part. The correct part is arithmetic: birthdays supply only days 1–31 and months 1–12, a small slice of all Lotto 6/45 combinations. The unverified part is how many people actually pick that way. This page computes the structure exactly, shows what would happen if people did cluster as a what-if, and draws only the conclusions the real data allow. First, one fixed point: whatever numbers you choose, one ticket's jackpot chance is 1 in 8,145,060.
How small is the area birthdays cover?
The number of combinations where all six numbers are 31 or below is C(31,6) = 736,281, which is 9.04% of 8,145,060. The other 90.96% (7,408,779 combinations) contain at least one number from 32 to 45. Combinations using only numbers up to 12, as with months alone, number 924, or 0.011%.
| Range | Combinations | Share of all |
|---|---|---|
| All in 1–31 | 736,281 | 9.04% |
| All in 1–12 | 924 | 0.011% |
| At least one 32 or above | 7,408,779 | 90.96% |
| All combinations | 8,145,060 | 100% |
Birthdays usually fill only a few of the six numbers; a handful of family dates cannot fill a whole ticket, and the rest is chosen another way. So what share of real tickets use only numbers up to 31 is a separate figure, and this site does not know it.
What-if: how a share changes if people cluster
The next table is an assumption, not an observation. Suppose a fraction f of all selections is chosen uniformly among combinations with every number at or below 31, and the rest uniformly among all combinations. Then a given all-≤31 combination is chosen f/0.0904 + (1 − f) times as often as an average one, and any other combination (1 − f) times as often. Taking an average combination to produce 14 winners in a round (the same basis as the previous article), we compute the expected share of the pool if you win.
| Fraction f | Popularity of all-≤31 combination | Popularity of other combinations | Share, all-≤31 | Share, others |
|---|---|---|---|---|
| 0% | 1.00× | 1.00× | 7.14% | 7.14% |
| 10% | 2.01× | 0.90× | 3.56% | 7.94% |
| 20% | 3.01× | 0.80× | 2.37% | 8.93% |
| 30% | 4.02× | 0.70× | 1.78% | 10.20% |
If only 10% of selections cluster in the ≤31 area, a combination there is twice as popular and its expected share on winning is less than half. At 30% the gap is 5.7-fold. The table shows a structural fact: because the area is small, even a modest fraction of people makes it dense. What it cannot show is what fraction truly clusters.
You can change the assumptions in the tool below.
What the real data show
One piece of data is available: for each round, how many of the six winning numbers are 31 or below, set against observed ÷ expected first-prize winners. Results for the 1,160 rounds from 88 to 1247:
| Numbers ≤ 31 | Rounds | Ratio |
|---|---|---|
| 2 or fewer | 71 | small sample, unstable |
| 3 | 242 | 0.938 |
| 4 | 408 | 1.048 |
| 5 | 338 | 0.972 |
| All 6 | 101 | 1.070 |
All six numbers were 31 or below in 101 of 1,160 rounds, 8.7%, matching the expected 9.04%; the draw itself is not skewed toward low numbers. The ratios wander between 0.94 and 1.07 without a clear direction. A permutation test comparing rounds with five or more such numbers against rounds with three or fewer gives a one-sided p of about 0.05 (0.995 against 0.940): the same direction as the hypothesis, but weak evidence. And round 1152 (30, 31, 32, 35, 36, 37), with few numbers at or below 31, also drew many winners. This site does not present "birthdays cause the clustering" as established.
Clustering exists beyond birthdays
Clustering is unlikely to come from birthdays alone. What follows is a list of possible reasons, not facts this site has verified.
- A tendency to pick numbers that sit next to each other on the play slip, or in lines (rounds 1152 and 1162 look like adjacent-number picks).
- Following numbers considered lucky, last week's winning numbers or so-called hot numbers (the hot-number article found these have no predictive power).
- The habit of repeating the same numbers every week.
We could not separate which of these produces real clustering. That unusual rounds share a high manual share is clear (83% in round 1128), but the reasons within it are unknown.
Questions that come up
Is it a mistake to play birthday numbers, then? Not as far as the chance of winning goes: every choice is 1 in 8,145,060. If there is a downside it is only the share after a win, if clustering truly exists, and that is an assumption here, not something this site has verified.
What if only some of my numbers are birthdays? Most people probably use a few birthdays and choose the rest another way. The what-if table applies only when all six numbers are 31 or below. If only some are, the overlapping area is much wider and the clustering effect much weaker. Read the table as "the extreme case of six birthday numbers".
Are numbers up to 12 riskier because of months? In the data the ratio rose slightly with more numbers at or below 12 (1.19 for four numbers in 36 rounds, 1.58 for five in 5 rounds), but the samples are tiny. This page does not rely on those figures.
Does the number generator deliberately include numbers above 31? No. The generator on this site draws every number with equal probability, so combinations of 31 or below and the rest appear in proportion to their share of all combinations.
Is there a way to check for clustering? The most direct way is to collect the number shapes of rounds with many winners. The permutation test here is one such attempt, and the result, "plausible but not established", shows the limits of the material available today.
How to read the what-if table
Every number in the what-if table moves when the clustering fraction f changes. What to remember from it is therefore the direction rather than any single figure: the larger f is, the faster the share of ≤31 combinations falls, while the share of the others rises slightly.
"So should I pick numbers above 31?"
That is not the right reading. Precisely, this page says:
- Every choice of numbers has the same chance of winning. Choosing numbers above 31 does not raise your chance of winning at all.
- If clustering is real, the only thing affected is the share you would receive if you win: a matter that arises only after an extremely improbable event.
- Expected return is negative for any choice. A larger share does not change that structure.
- Auto selection, or this site's generator, which draws numbers at random, does not push them toward any area. That does not improve the chance of winning; it only means no clustering bias.
One more realistic remark. With a 1 in 8,145,060 chance per ticket, a clustering-driven difference in shares matters only when a rare event occurs. Separating probability from share is a habit that applies beyond lotteries: with equal odds, the way an outcome is divided still changes the value one can expect.
LottoFlick is independent and not an official lottery operator. Numbers are random and cannot be predicted.
Sources and verification
- LottoFlick aggregate output: winners by how many drawn numbers are 31 or below, permutation test — computed by LottoFlick
- Donghaeng Lottery: Lotto 6/45 round results — verified