Why a Shared Birthday Is Likely Among Just 23 People

Assume a 365-day year, ignore 29 February, and treat each person’s birthday as independent and uniformly distributed. Among 23 people, the probability that at least two share a birthday is about 50.7 per cent. This is not asking whether someone is likely to meet another person with their own birthday. It asks whether any pair in the group matches.

The simplest calculation begins with the opposite event: no shared birthdays at all. The first birthday is unrestricted. The second must avoid one date, giving a probability of 364/365; the third must avoid two dates, giving 363/365; and so on, down to 343/365 for the twenty-third person. Multiplying these values gives about 49.3 per cent. Subtracting that from one gives roughly 50.7 per cent for at least one match.

What intuition often misses is the number of pairs. Twenty-three people do not create merely 23 opportunities for a match; they create 23 × 22 ÷ 2 = 253 pairs. The group grows modestly, but its pairwise relationships grow much faster, so coincidences accumulate quickly.

The “birthday paradox” is not a logical contradiction. It is a gap between intuition and calculation. Real birthdays are not perfectly uniform, and relatives or same-age groups may not be independent, so 50.7 per cent belongs to a simplified model. The model nevertheless reveals the key structure: to judge whether a group contains a match, count every possible pair.

https://mathworld.wolfram.com/BirthdayProblem.html
https://www.stat.berkeley.edu/~aldous/Top_Ten/talk.html