Paper is ruled with lines a distance t apart, a needle of length L ≤ t is dropped, and the question is the probability that it crosses a line. The answer is 2L/(πt): with the needle as long as the gap, 2/π = 0.63661977236758, the first link; with a needle half as long, 1/π = 0.31830988618379, the second. The π is there because the needle's angle is uniform and the crossing condition involves its sine, and the average of sin θ over a half-turn is 2/π — which is the whole derivation in one sentence, and the reason the constant cannot be legislated away.
Laplace noticed that the argument runs backwards: throw N needles, count the crossings, and π becomes the unknown in a measurement rather than a constant in a proof. This is the first Monte Carlo calculation, and it is a bad one. The scatter of a count of N throws is of the order of √N, so the relative error in 2/π is about 0.76/√N and six correct digits of π would cost something like half a trillion throws — eighteen thousand years of dropping a needle once a second — and would be beaten by any series a schoolchild can write down. The generalisation is worth more than the experiment: Barbier's theorem, which falls out of the same linearity, says that any curve of length L has expected crossings 2L/(πt), so a circle whose diameter is the gap crosses exactly twice whatever its orientation, which is a proof that its circumference is πt and that π is the same number in every one of these questions.
Answer
0.63661977236758
How to read it
2/π on the workbench, the crossing probability for a needle as long as the line spacing. The second link is a needle half as long.
Source
Georges-Louis Leclerc de Buffon, 1733, on the probability that a thrown needle meets one of a set of parallel lines; the formula 2L/(πt) and Laplace's observation that the experiment therefore estimates π
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.