applets.

Galton Board

← All apps

About Galton Board

Steel balls fall through rows of brass pins. At every pin a ball goes left or right, and the bins at the bottom fill up in the shape of a bell. Pick how many balls and how fast, then press DROP; more balls land on top of the ones already there until you press CLEAR. Choose 1 to drop a single ball and trace its path (L for left, R for right). Point at a bin (or tap it) to see how many balls it caught and the chance of a ball landing there, as "1 in N". Rows sets the number of pins a ball meets; Lean tilts every pin so a ball goes right more (or less) often than half the time. Pascal writes on each pin how many different paths lead to it.

Keys: Space drop or stop, C clear, P Pascal's triangle; with the board selected (Tab to it), ← / → read each bin. COPY LINK copies a link to this setup.

THE STORY

Francis Galton, Charles Darwin's half-cousin, had this machine built in 1873 to show how the bell-shaped "law of frequency of error" comes about. He called it the quincunx, after the pattern its pins are set in: each row sits in the gaps of the row above, like the five spots on a die repeated over and over. He described it in his book Natural Inheritance (1889), where he also imagined a two-stage version to explain why the children of very tall parents tend to be less tall than their parents: what he called regression. In the same book he wrote: "I know of scarcely anything so apt to impress the imagination as the wonderful form of cosmic order expressed by the 'Law of Frequency of Error'."

The maths is older. Blaise Pascal's Treatise on the Arithmetical Triangle (written in 1654, published 1665) set out the triangle of numbers now named after him, though it was known centuries earlier in Persia, India and China, where it is called Yang Hui's triangle after a book of 1261. In 1733 Abraham de Moivre showed that for many coin tosses the binomial chances follow a smooth bell curve, the one now called the normal curve.

HOW IT WORKS

With n rows a ball makes n left-or-right choices, so there are 2n possible paths. It ends in bin k if exactly k of its bounces go right, and the number of paths that do that is "n choose k", the number in row n of Pascal's triangle. With level pins every path is equally likely, so a bin's chance is its number of paths ÷ 2n; with a lean p each path has chance pk(1 − p)n−k. That is the binomial distribution, drawn in red as Expected: what the balls should settle at. Its average is n·p and its spread (standard deviation) √(n·p·(1 − p)). The dashed blue Normal curve is de Moivre's smooth approximation with the same average and spread; it fits better the more rows there are, and worse when the lean is strong.

Each ball's path is chosen at random as it is dropped; the bouncing you see is drawn to follow that path, falling under a steady gravity between pins. A real board is never quite this fair: balls carry speed from one pin to the next and are a little more likely to keep going the same way, which makes a real bell somewhat wider. As the bins fill, the balls are drawn smaller so the piles always fit; past a few thousand they pour like sand. Balls in a bin settle in close-packed rows.

FURTHER WATCHING AND READING

BALLS IN BINS0 
AVERAGE BIN– 
SPREAD– 
STILL TO FALL0 

The Drop

BALLS
SPEED

The Board

ROWS12
LEANlevel
SHOW

The Bins

BinBallsShareExpectedChance

Why a Bell?