Method 5 · 1777

Dropping needles on a floor

No circles. No fractions. Just needles, a lined floor, and a count. Out pops π. Nobody saw this coming.

The setup

Take a floor with parallel lines drawn on it, evenly spaced. Now grab a pile of needles, all the same length as the line spacing, and drop them one at a time. Where each needle lands, and how it's rotated, is completely random.

Some needles land lying entirely between two lines. Others happen to cross a line. Count both. Then divide twice the total number of needles by the number of crossings. That number is π.

Why this works

Whether a needle crosses a line depends on two random things: how far its center is from the nearest line, and how much it's tilted. With a bit of geometry and an integral over all possible angles, the probability of crossing comes out to exactly 2 / π when needle length equals line spacing.

That π in the denominator is wild. There's no circle anywhere in the setup. It sneaks in through the average projected length of a randomly oriented line — which involves an integral of |sin θ| — and that integral is where π comes from. Trigonometry's relationship with circles is doing the smuggling.

If P (crossing) = 2 / π, then π = 2 / P. We don't know P directly, but we estimate it from the data: crossings / total. Substitute and you get π ≈ 2 × total / crossings.

Buffon, 1777

Georges-Louis Leclerc, the Comte de Buffon, posed this problem in 1777 as part of a broader inquiry into geometric probability. He was a naturalist and polymath as much as a mathematician — best known for his massive Histoire Naturelle. The needle problem was a side thought that became one of the most famous demonstrations in probability theory.

It is, to be clear, a terrible way to actually compute π. It converges as 1 / √N, just like Monte Carlo: 100 needles for one digit, 10,000 for two, a million for three. But unlike Monte Carlo, it has the peculiar charm of physical randomness — you can do this with real needles in your kitchen, and people have. Mario Lazzarini in 1901 famously claimed to get π = 3.1415929 from 3,408 needle tosses, an accuracy that's almost certainly the result of stopping the trial at a convenient moment, not honest data. (See: every statistics textbook's cautionary tale.)

Try this

• Drop 50 needles and check the estimate. Then 500. Then auto-drop for a while.
• Note how unstable the estimate is at low needle counts — π is in there, but it takes thousands of drops to lock in even one digit reliably.
• If you have actual needles and a lined notebook, try it for real. Just don't stop early to "improve" your answer.

Sources & further reading

• Wikipedia: Buffon's needle problem — the math, including the integral derivation.
• Wikipedia: Comte de Buffon — his much broader naturalist career.
• Wikipedia: Geometric probability — the field Buffon's needle helped establish.