The birthday problem vs the lottery: why coincidences are cheap and jackpots aren't

Put 23 random people in a room and there is a better-than-even chance two share a birthday. Most people guess you'd need 180-odd. The gap between the guess and the truth is the same gap that makes "someone always wins the lottery" feel like evidence that winning is achievable. Same illusion, same fix.

The birthday problem, worked

Assume 365 equally likely birthdays (ignore 29 February; it barely moves the answer). Compute the probability that nobody shares, then subtract from 1.

Person 1 takes some day. Person 2 avoids it with probability 364/365. Person 3 avoids both with 363/365, and so on. For 23 people:

P(no shared birthday) = (365/365) × (364/365) × (363/365) × … × (343/365)

Multiply the 23 factors out and you get 0.4927. So:

P(at least one shared birthday) = 1 − 0.4927 = 0.5073 → 50.7%

— the classic result you'll find derived the same way at Wolfram MathWorld. With 22 people it's 47.6%, so 23 is exactly where the coin tips.

Why 23 is enough: count the pairs, not the people

The intuition repair is to notice what's actually being tried. A shared birthday needs only some pair to collide, and 23 people contain:

C(23,2) = 23 × 22 ÷ 2 = 253 pairs

Each pair collides with probability 1/365. As a back-of-envelope: the chance that all 253 pairs miss is about (364/365)^253 = 0.4995, giving roughly a 50% chance of a hit — pleasingly close to the exact 50.7%. The paradox dissolves: 253 lottery-style attempts at a 1-in-365 target is not a long shot at all.

Contrast that with the question people think is being asked: "does someone here share my birthday?" Now only 22 comparisons matter, each 1/365:

P(nobody matches me) = (364/365)^22 = 0.9415 → only a 5.9% chance of a match

Some-pair versus this-specific-pair: 50.7% against 5.9%, from the same room of people. Specificity is expensive.

The lottery is the birthday problem wearing a jackpot

Swap birthdays for lottery lines and the two questions reappear:

• "Will someone win?" is the some-pair question. When a big US Powerball draw sells, say, 200 million tickets against 292,201,338 combinations, the expected number of quick-pick jackpot hits is 200,000,000 ÷ 292,201,338 = 0.68 — winners are routine, and rollovers still happen regularly, which is exactly what that number predicts. Sell enough attempts and somebody colliding with the drawn numbers is unremarkable, just like the 253 pairs.
• "Will I win?" is the my-birthday question, except radically worse: 1 in 292,201,338 per line (official odds) instead of 1 in 365 per pair.

So "someone always wins eventually" is true, boring, and utterly uninformative about your ticket. The news shows you the collision that happened; it cannot show you the 292 million tickets' worth of nothing that funded it. Treating a visible winner as evidence of winnability is called survivorship reasoning, and lotteries are its perfect habitat — the same effect that makes double winners inevitable across a big enough population (the odds of winning twice).

One more birthday-flavoured lottery fact

The birthday problem also predicts a genuinely practical lottery effect: collisions between players. Millions of players choosing from the same small pool of "lucky" numbers (dates cap picks at 31) means popular lines are shared by hundreds of people, and a pari-mutuel jackpot splits accordingly. Coincidence being cheap works against you here too — priced out in the number sharing risk tool.

The takeaway rule

Whenever a probability claim surprises you, ask: how many attempts does the universe get? Coincidences look miraculous when you count people (23) instead of attempts (253). Lotteries look winnable when you count winners (1, on the news) instead of tickets (hundreds of millions, invisible). One number is the room; the other is the raffle. For what the raffle really looks like from inside one ticket, see picturing 1 in 292 million — or run a few thousand losing weeks in the draw simulator and watch coincidence stay cheap for everyone but you.