🎲 Birthday Paradox Calculator
Find the real odds that two people in your class, team, office or party share a birthday — and see why the answer is far higher than almost anyone guesses.
🔍 Reverse it: how many people do you need?
Pick a confidence level and this tells you the smallest group size that reaches it.
📖 How It Works
The birthday paradox is not really a paradox — it is a counting problem that our intuition gets badly wrong. Most people answer "23 people? The odds must be tiny — 23 divided by 365 is about 6%." That reasoning has a hidden flaw: it counts only the comparisons against your own birthday, while the real question counts every possible pair in the group.
With 23 people there are 23 × 22 ÷ 2 = 253 unique pairs. Each pair is a fresh chance of a match, and 253 shots at a 1-in-365 coincidence is a lot more than most people expect. That is why the group size needed for a 50/50 chance is only 23, not 183.
The clean way to compute it is by working backwards. Instead of adding up the probability of a match (which gets messy because matches can overlap), calculate the chance that everyone is different — an easy chain of multiplications — then subtract it from 1.
P(shared birthday) = 1 − P(all different)
Swap 365 for 366 and the answers barely move. February 29 actually makes a real-world match slightly less likely than the idealised model, because leap-day birthdays are rarer than 1 in 366 in practice — but the difference is under half a percentage point at any realistic group size.
| People in group | Pairs compared | Chance of a shared birthday (365 days) | Chance (366 days) |
|---|---|---|---|
| 5 | 10 | 2.7% | 2.7% |
| 10 | 45 | 11.7% | 11.7% |
| 15 | 105 | 25.3% | 25.2% |
| 20 | 190 | 41.1% | 41.1% |
| 23 | 253 | 50.7% | 50.6% |
| 25 | 300 | 56.9% | 56.8% |
| 30 | 435 | 70.6% | 70.5% |
| 35 | 595 | 81.4% | 81.3% |
| 40 | 780 | 89.1% | 89.1% |
| 41 | 820 | 90.3% | 90.3% |
| 45 | 990 | 94.1% | 94.0% |
| 50 | 1,225 | 97.0% | 97.0% |
| 57 | 1,596 | 99.0% | 99.0% |
| 60 | 1,770 | 99.4% | 99.4% |
| 70 | 2,415 | 99.92% | 99.92% |
Figures assume birthdays are spread evenly across the year. Real birth data has seasonal peaks (late summer in the US), which pushes the true odds very slightly higher than the table above.
💡 Why It Matters
🎉 Planning a party game
A "find your birthday twin" icebreaker sounds like a dud with 25 guests — but the odds are actually 57%, so it lands more often than not. Knowing the numbers stops you betting on the wrong side of a coin flip.
🏫 Teachers and coaches
In a 30-student class there is a 71% chance of a shared birthday, which is why school "birthday charts" and team celebrations need a plan for double-ups. Pick 23 students at random and it is already a coin flip.
🧑💼 HR and team celebrations
An office of 50 colleagues has a 97% chance two people share a birthday. If your team calendar has only one cake a month, it is not unusual — it is the default outcome of a group that size or larger.
📈 Learning real probability
The birthday problem is the classic teaching example of why we should compute rather than trust intuition. The same "count the pairs, not the people" trick explains coin-flip streaks and hash collisions in computing.
❓ Frequently Asked Questions
Why is it called a paradox if the maths is simple?
It is called a paradox because the true answer conflicts so sharply with intuition. Almost everyone predicts a much larger group is needed, so the correct figure of 23 people for a 50/50 chance feels impossible even after the calculation is explained.
What is the smallest group with a guaranteed match?
A guaranteed match is impossible until you have 367 people, because there are only 366 possible birthdays. By that point the pigeonhole principle forces a duplicate, whether or not anyone intends it — no probability required at all.
Does the calculator count my own birthday specifically?
No — and this is the most common mistake. The odds that someone shares your birthday are much lower: it takes around 253 people for a 50/50 chance, which is about 11 times more people than for any match at all.
Do leap-year birthdays change the result much?
Only marginally. Moving from 365 to 366 possible dates shifts the 50% mark from 23 to about 24 people and changes most probabilities by less than a percentage point, so most published tables ignore it entirely.
Are birthdays really spread evenly across the year?
No. Real birth records show clear seasonal clusters, with peaks in late summer and dips on major holidays. Uneven distribution makes matches slightly more likely than the even model, so the table is a mild under-estimate.