The Birthday Paradox
The collision probability crosses 50% at 23 people because the chance that everyone is unique is the product of shrinking 'not-a-match' probabilities, and that product falls below 0.5 much sooner than a count intuition suggests.
A complete interactive classroom, not just a preview.
Start when you are ready to enter this Stage's 8 scenes and explore, respond, and learn as you go.
How many people do you need in a room before it's more likely than not that two share a birthday?
Most people guess you'd need 183 or 365 people — but the real answer is shocking and changes how you think about probability.
Your intuition says birthdays should rarely collide, yet a small group can make a collision almost certain. Where does that intuition go wrong?
An interactive probability calculator where learners add people one at one and watch the collision probability climb from 0% to over 50%, then visualize all birthday pairs in a grid.
Just 23 people are enough to make it more likely than not that two share a birthday — and the reason is a cascade of 'not-colliding' chances that shrinks fast.
You probably need around half of 365, roughly 183 people, since each new person has to 'miss' all previous birthdays.
- Specific date or zodiac analysis
- Leap year adjustments beyond a brief mention
- Variants like 'three people sharing a birthday'
- Historical origin of the problem
- 01The Provocative QuestionslideQuestion
Open with the classic puzzle and surface the learner's gut answer before any math appears.
- Two people in a room — could they share a birthday?
- How many people until a match becomes likely?
- Surface the learner's first guess
- 02Commit to Your GuessquizPrediction
Before seeing any math, the learner commits to a number of people needed for a 50%+ chance of a shared birthday.
- Pick one answer from a short list
- There is no penalty for guessing wrong — the answer is calibrated to feel surprising
- 03Watch the Probability BuildinteractiveEvidence
A simulator where the learner adds people one at a time and watches the probability of at least one shared birthday update live.
- Each new person multiplies by a smaller chance of being unique
- Crosses 50% at 23 people
- Crosses 99% at just 57 people
- 04See All Possible PairsinteractiveEvidence
A grid visualization where every dot represents one possible pair of people; pairs that share a birthday light up.
- With n people there are n(n−1)/2 pairs
- Even a small n produces many opportunities for a collision
- Concrete visual reason why collisions arrive fast
- 05Why 23 Is the Magic NumberslideExplanation
Walk through the formula: chance everyone is unique equals the product of (365−k+1)/365 for k from 1 to n, and that product dips below 0.5 at n=23.
- P(unique) = 365/365 × 364/365 × 363/365 × …
- Each fraction is just below 1, but they multiply
- Logarithms show the product halves roughly every 6 people
- 06Where the Model BreaksslideBoundary
Clarify assumptions: 365 equally likely days, ignoring leap years and real-world clustering effects.
- Assumes uniform distribution across 365 days
- Real births cluster (more Sept babies, fewer Feb 29)
- Real-world curves differ slightly but the 23-person threshold is robust
- 07Try It in a New SettinginteractiveTransfer
Let the learner test a changed scenario: what if birthdays were 1 in 100 instead of 1 in 365? How many people then?
- Slider changes the number of possible 'days'
- Threshold scales with the square root of possibilities
- Connects the birthday problem to hash collisions and the pigeonhole principle
- 08The Answer, and Why It MattersslideResolution
Close by directly answering the driving question and connecting the result to how probability behaves in surprising ways.
- 23 people — that's the answer
- Intuition misjudged growth rates, not the math
- Same structure appears in cryptography, polling, and the pigeonhole principle
Discussion threads for a Stage aren't available yet.
