On the Galton board, every ball bounces left or right at every peg by pure coin-flip, totally unpredictable. But drop lots of balls and a smooth hill always grows in the middle. Chance is unpredictable one ball at a time, and dead reliable in crowds.
Most people think randomness should spread balls out evenly. In fact the centre bin is reachable by thousands of left/right routes and each edge by only one, so a smooth bell curve rises every single time.
What's actually happening
The Galton board is a machine for confessing what randomness really is. Watch one ball: nine coin flips, a drunken stagger, an unguessable destination. Watch a hundred and fifty: a smooth, symmetrical hill rises in the centre bins, every single time, with edges thinning out exactly on schedule. Nothing about ball #73 became predictable. The crowd did. That is probability's whole bargain: it gives up on individuals and becomes precise about populations.
The hill's shape has a reason. To land in the middle bin, a ball needs roughly equal lefts and rights, and there are many different flip-sequences that do that (LRLRLRLRL, RRLLRLLRL…). To land at the far edge it needs all nine flips to agree, and there is exactly one sequence for that. The bins simply count routes: middle destinations are reachable by thousands of paths, extremes by one. Pile up the route-counts and you get the binomial distribution, which smooths into the bell curve as the rows increase.
Here is why this one toy explains so much of the world: the bell curve appears wherever an outcome is the sum of many small, independent accidents. Your height (thousands of genes plus nutrition), a measurement's error (dozens of tiny disturbances), a poll's noise: each is a ball falling through its own pegboard. Francis Galton, who built the first board in 1874, called the effect "the supreme law of unreason": individually lawless events, collectively lawful. It's why casinos always profit and insurers can promise: they never know the next ball, and never need to.
Chance is unpredictable for one ball and dead reliable for a crowd — the bell curve appears wherever many small accidents add up, which is why casinos always profit.
- 1Flip a real coin 50 times, recording H/T. Separately, write down a fake "random" 50-flip sequence by hand, trying to make it look real.
- 2Now count the longest run of identical results in each. Hand-faked sequences almost never dare a run longer than 3; real chance usually produces a run of 5 or 6.
- 3Show both to a friend and ask which is real. Genuine randomness is streakier than anyone believes — which is also why hot streaks in games feel so meaningful and usually aren't.
Common questions
Many different flip-sequences land a ball in the centre bin, but only one sequence reaches each far edge. The bins simply count routes, and the middle is reachable by thousands of paths.
It shows up wherever an outcome is the sum of many small, independent accidents — height, measurement error, polling noise. Each is a ball falling through its own pegboard. This is the central limit theorem.
A roulette wheel is random per spin but a metronome per million spins. The house edge is invisible to any one player and inevitable across the crowd — the bell curve pays the bills.