This birthday paradox calculator computes the probability that at least two people in a group share a birthday — the famously counterintuitive result that just 23 people make it more likely than not (50.7%).

–

253 Pairs Hide Inside 23 People

The birthday paradox is not a paradox of probability but of phrasing: asked "what's the chance someone shares MY birthday," the honest answer for 22 strangers is small (~5.9%); asked "what's the chance ANY two share A birthday," the counting changes completely because every pair is a ticket. Twenty-three people hold C(23,2) = 253 pairs, and the survival probability that none of them collide — (365/365)(364/365)(363/365)…(343/365) — decays to 0.4927, leaving the famous 50.73% for at least one match. The curve then sprints: 30 people → 70.6%, 40 → 89.1%, 50 → 97.0%, 57 → 99.0%, 70 → 99.92%. The approximation P ≈ 1 − e^(−n²/730) tracks the exact product closely enough for any whiteboard, and it exposes the general law underneath: collisions in a space of size d arrive around n ≈ √d — the "birthday bound" that cryptographers invoke when they say a 128-bit hash offers only 64-bit collision security.

The assumptions deserve their footnote, because reality is slightly more generous than the model. The calculation assumes birthdays uniform over 365 days (Feb 29 excluded — include it by setting the days field to 366 and the odds dip a hair) and independent. Real births are neither: they cluster seasonally (more September birthdays in northern-hemisphere data) and spike after holidays, and non-uniformity can only RAISE collision odds (any concentration makes matches likelier). Twins in the room, obviously, cheat outright. So 50.7% at 23 people is a floor for real-world groups, which is why the bar trick works even better in practice — and why the same math, run on your class, office or football squad, keeps "surprising" people whose intuition was answering the wrong question all along.

Worked Examples (Verified)

  • Defaults: 23 people → 50.73%.
  • A classroom of 30 → 70.63%.
  • An office of 50 → 97.04%.
  • Just 10 people → 11.69% — pairs (45) are already working.
  • Leap-inclusive year (366 days): 23 people → 50.63% — one extra day costs a tenth of a point.

Birthday Paradox Discipline

  • "Shares YOUR birthday" and "shares A birthday" are different events with wildly different odds — most confusion is the question, not the math.
  • Uniformity is the conservative assumption; real birthday clustering pushes true odds slightly UP.
  • The product form is exact; the e^(−n²/2d) shortcut is an approximation that degrades gracefully (fine to n ≈ 100 at d = 365).
  • Generalize carefully: the same formula prices hash collisions, duplicate random IDs, and DNA-profile matches — with d set to your actual space size.
  • Pigeonhole caps the story: n = d + 1 (366 people, 365 days) makes a match certain, not just likely.

Related Tools

Frequently Asked Questions

How many people do you need for a 50% chance of a shared birthday?

Twenty-three. With 23 people there are 253 possible pairs, and the probability that at least one pair shares a birthday is 50.73% — computed as 1 minus the product (365/365)(364/365)…(343/365). This calculator gives the exact probability for any group size (and any number of possible birthdays, so you can test 366 for leap years or invent other calendars).

Why is the birthday paradox so counterintuitive?

Because intuition answers a different question. "Someone shares MY birthday" stays unlikely until the group approaches ~253 people. "ANY two share A birthday" counts every pair — and pairs multiply quadratically: 23 people = 253 chances, 50 people = 1,225 chances. The surprise is the gap between linear intuition (more people, proportionally more chance) and quadratic reality (more people, squared chances). Nothing about the probability is paradoxical once the event is stated precisely — which is the general lesson the problem exists to teach.

What is the probability that two people share a birthday in a group of 30?

70.63%. A class of 30 — the size teachers actually face — is already a better-than-2-in-3 bet, and real classrooms tilt further because births cluster seasonally rather than uniformly. For reference: 20 people → 41.1%, 40 → 89.1%, 50 → 97.0%, 60 → 99.4%. The calculator reproduces the full table for your group, and entering 366 days shows the (slightly lower) leap-year-aware figures.

How does the birthday paradox relate to computer security?

Directly — it IS the birthday attack. Finding two inputs with the same hash output (a collision) takes about √d tries in a space of d outputs, not d/2: a 128-bit hash falls to ~2⁶⁴ work, not 2¹²⁷. That's why security levels are quoted at half the hash length, why 256-bit hashes are the standard for 128-bit collision security, and why database systems size random ID spaces by this same formula. Swap "people" for random values and "birthdays" for possible outputs and this calculator is pricing collisions in your system.

What if we want two people to share a SPECIFIC birthday?

Then the quadratic magic vanishes: the probability that at least one of n people was born on a given date (say, yours) is 1 − (364/365)ⁿ — for 23 people just 6.1%, and you need ~253 people for a coin flip. The contrast between 23 (any match) and 253 (your match) is the cleanest demonstration that the paradox is about PAIRS, and it's the version to reach for when someone at the party insists the 23-person figure must be wrong.

Are birthdays really uniformly distributed?

Not exactly — real birth data shows seasonal waves (late-summer/autumn peaks in many countries), weekday effects in scheduled deliveries, and small spikes and dips around holidays. Every departure from uniformity increases collision probability (concentration can only help matches), so the uniform assumption in this calculator understates real-world odds slightly. For the classic question the effect is a percentage point or less at classroom sizes; for precise work, empirical birth distributions exist — and twins, of course, are a guaranteed collision walking in the door.

What group size makes a shared birthday 99% certain?

57 people reaches 99.0%, and 70 reaches 99.92% — after which the remaining probability mass is a rounding error with a long tail: certainty strictly requires 366 people (pigeonhole principle with 365 days), but practical certainty arrives far earlier. The steep middle of the curve (23 → 57 people carries you from 50% to 99%) is the quadratic signature again: each additional person adds more pairs than the last, so probability compounds much faster than headcount.

Is this birthday paradox calculator free?

Yes — free, instant, no sign-up. Enter the group size (and days in the year if not 365) for the exact shared-birthday probability.