Ask 23 strangers and the chance that two of them share a birthday is 0.50729723432399 — the workbench gets there as 1 − permutations(365,23)/365²³, where the numerator is the count of ways all 23 birthdays can come out different. It passes half by a whisker, which is why the result still reads as a paradox. Fifty-seven people push it past 0.99.
The counterintuition comes from counting people instead of comparisons: 23 people make combinations(23,2) = 253 pairs, and 253 chances at about 1/365 each is what crosses the half-way line, not 23. The same arithmetic is why hash collisions arrive far earlier than the table size suggests — the square root of the space is where trouble starts, which is why 128-bit hashes are treated as offering about 64 bits of collision resistance.
Answer
0.50729723432399
How to read it
1 − permutations(365,23)/365²³ on the workbench. The second link gives the complementary 0.49270276567601; both are float values, so the last digits are rounding, not arithmetic.
Source
The birthday problem, popularised by Richard von Mises in 1939; the 365-day uniform model ignores leap days and the real seasonality of births
Work it in the calculator
These links open the right calculator with this problem's numbers already typed in. The line above is what the screen should show — check it yourself.