Buffon's Needle: Estimating π by Dropping Toothpicks

In 1733 a French naturalist named Georges-Louis Leclerc, Comte de Buffon, asked a question that sounds more like a parlor game than mathematics. Take a floor made of evenly spaced parallel lines — floorboards, say. Drop a needle on it at random. What is the chance the needle comes to rest crossing one of the lines?

The answer, which Buffon published in 1777, contains a shock. The probability is

$P = \frac{2L}{\pi d},$

where $L$ is the length of the needle and $d$ is the spacing between the lines (for a needle no longer than the gap). The circle constant $\pi$ has appeared in a problem with no circles anywhere in sight — just straight lines and a straight needle. Rearranging the formula turns the experiment into a machine for computing $\pi$ : drop the needle many times, measure how often it crosses, and

$\pi \approx \frac{2L}{d} \cdot \frac{\text{number of throws}}{\text{number of crossings}}.$

This is the founding problem of geometric probability, and the ancestor of one of the most important computational techniques of the modern era. This article explains why it works.

Setting up the experiment

To pin the problem down, we need to describe where a dropped needle lands. Two random quantities determine everything. The first is the distance $y$ from the center of the needle to the nearest line below it; since the lines are spaced $d$ apart, $y$ ranges from $0$ to $d/2$ . The second is the angle $\theta$ the needle makes with the lines, ranging from $0$ to $\pi/2$ (a needle tilted by more than that is just a mirror image of one tilted by less).

A random drop means $y$ and $\theta$ are each uniformly distributed over their ranges, independently of one another. The picture below shows several needles on a lined floor: some cross a line, some sit entirely between two lines.

When does the needle cross?

Consider a needle of length $L$ tilted at angle $\theta$ . Its center is at distance $y$ from the nearest line. The needle reaches a vertical distance of $\frac{L}{2}\sin\theta$ above and below its center. It crosses the line precisely when that vertical reach is at least as large as the gap to the line:

$y \le \frac{L}{2}\sin\theta.$

So for a given tilt $\theta$ , the favorable values of $y$ are those between $0$ and $\frac{L}{2}\sin\theta$ . The key feature is the $\sin\theta$ : a needle lying nearly parallel to the lines ( $\theta \approx 0$ ) almost never crosses, while a needle nearly perpendicular ( $\theta \approx \pi/2$ ) crosses unless its center sits right in the middle of a board. The sine function interpolates smoothly between these extremes — and it is the sine that will summon $\pi$ .

Doing the integral

The probability is the favorable area divided by the total area in the space of possibilities $(\theta, y)$ . The total region is the rectangle $0 \le \theta \le \pi/2$ and $0 \le y \le d/2$ , with area $\frac{\pi}{2}\cdot\frac{d}{2} = \frac{\pi d}{4}$ .

The favorable region is everything under the curve $y = \frac{L}{2}\sin\theta$ . Its area is the integral

$\int_0^{\pi/2} \frac{L}{2}\sin\theta \; d\theta = \frac{L}{2}\Big[-\cos\theta\Big]_0^{\pi/2} = \frac{L}{2}\,(0 - (-1)) = \frac{L}{2}.$

The probability of crossing is the ratio of these two areas:

$P = \frac{\text{favorable area}}{\text{total area}} = \frac{L/2}{\pi d/4} = \frac{2L}{\pi d}.$

There it is. The $\pi$ entered through the width of the angle range — a full quarter-turn measured in radians is $\pi/2$ — while the sine integral contributed the clean factor of 1. The circle was never on the floor; it was in the set of directions the needle could point. Spinning through every orientation is, geometrically, a trip around a circle, and that is where $\pi$ lives.

Turning it into a π-machine

Because $P = 2L/(\pi d)$ , a single rearrangement isolates $\pi$ :

$\pi = \frac{2L}{d \cdot P}.$

In an experiment, the true probability $P$ is unknown, but the fraction of crossings approximates it. If you throw the needle $N$ times and it crosses $C$ times, then $P \approx C/N$ , so

$\pi \approx \frac{2L \, N}{d \, C}.$

Choose a needle exactly as long as the line spacing ( $L = d$ ) and this simplifies to the memorable $\pi \approx 2N/C$ — twice the number of throws divided by the number of crossings. Drop a toothpick on a tiled floor enough times, keep a tally, and the digits of $\pi$ slowly emerge from pure randomness.

The word “slowly” is doing real work. The accuracy of this estimate improves only like $1/\sqrt{N}$ : to gain one more reliable decimal place you need roughly a hundred times as many throws. Getting $\pi$ to three decimals would take on the order of a million careful drops. Several 19th-century experimenters reported needle estimates suspiciously close to $\pi$ — the Italian Lazzarini famously claimed $3.1415929$ from 3408 throws in 1901 — and statisticians now strongly suspect those results were stopped, or selected, exactly when the running estimate looked best. As a practical calculator, Buffon’s needle is a wonderful failure.

The real legacy

Its importance is not as a way to compute $\pi$ but as the first example of an idea that would become enormous: computing a fixed quantity by averaging random outcomes. That is the essence of the Monte Carlo method, developed by Stanislaw Ulam and John von Neumann at Los Alamos in the 1940s to model neutron diffusion, and named after the famous casino. When a quantity is too tangled to write down with a clean formula, you can often set up a random process whose average is that quantity, then sample it many times. The estimate’s error shrinks like $1/\sqrt{N}$ — exactly the rate Buffon’s needle exhibits — regardless of how many dimensions the problem has, which is precisely why Monte Carlo methods dominate high-dimensional calculations where ordinary numerical integration is hopeless.

Today Monte Carlo simulation prices financial derivatives, renders the soft lighting in animated films, estimates the integrals of quantum field theory, and powers a large slice of modern statistics. All of it traces back to a naturalist in the 1700s wondering how often a dropped needle would land across a crack in the floor — and being delighted to find $\pi$ staring back at him.

Frequently asked

Who was Buffon and when did he pose this problem?

Georges-Louis Leclerc, Comte de Buffon, was an 18th-century French naturalist best known for his vast encyclopedia of natural history. He posed the needle problem in 1733 and published the solution in 1777. It is regarded as the first problem in geometric probability — the branch of probability concerned with lengths, areas, and angles rather than with discrete counting.

Why does π appear in an experiment with no circles?

The hidden circle is the spinning needle itself. As the needle can land at any angle, its orientation sweeps through a full range of directions — and integrating over all those angles brings in a sine function, whose integral over a half-turn produces the factor that involves π. The circle is the set of all possible orientations, not anything drawn on the floor.

Is dropping needles a good way to actually compute π?

No — it is charming but hopelessly inefficient. Because the estimate converges like 1/√N, getting π accurate to even three decimal places would require an enormous number of throws, and famous historical 'experiments' that hit π very precisely are now suspected of being fudged. Its real value is conceptual: it was the seed of the Monte Carlo method, the modern technique of computing hard quantities by random sampling.

What is the Monte Carlo method?

It is a family of algorithms that estimate quantities by generating random samples and averaging the outcomes — exactly what Buffon's needle does. Developed at Los Alamos in the 1940s for neutron-diffusion calculations, Monte Carlo methods are now everywhere: in physics, finance, computer graphics, and statistics, wherever a problem is too complex to solve with a clean formula.

Buffon's Needle: Estimating π by Dropping Toothpicks

Setting up the experiment

When does the needle cross?

Doing the integral

Turning it into a π-machine

The real legacy

Frequently asked

Continue reading

The Coupon Collector's Problem: Why the Last Sticker Is Always the Hardest

The Galton Board: Watching the Bell Curve Emerge from Pure Chance

Brownian Motion: The Mathematics of Pure Randomness in Time

Keep exploring