Your Brain Cannot Do Random
Take a sheet of paper and write down twenty coin flips — H or T, twenty of them — without touching a coin. Just make them up, as randomly as you can.
Do it before you read on. It only works once.
Now count the longest stretch where the same letter repeats. Two in a row, three, four?
If your longest run is two or three, your list is almost certainly invented, and it can be spotted from across a room. Almost nobody writes a run of five. Real coins write them constantly.
The exact size of the tell
This is not a hunch. Every possible run of twenty flips can be listed by a computer — there are 1,048,576 of them, which is 2 multiplied by itself twenty times, exactly the doubling from part 4. Counting how many contain a long run takes seconds.
| A real run of 20 flips | How often it contains this |
|---|---|
| a run of 3 or more | 97.9% |
| a run of 4 or more | 76.8% |
| a run of 5 or more | 45.8% |
| a run of 6 or more | 23.7% |
Read the top line again. Nearly ninety-eight real strips in every hundred contain three of the same in a row. A made-up strip with a longest run of two is not unusual — it is in the rarest 2%.
And a run of four turns up in more than three-quarters of real strips, while people making them up produce one about as often as they produce a run of nine.
There is a second tell too. People invent almost exactly ten heads and ten tails, because "half and half" feels like what random means. Real twenty-flip strips land on exactly ten heads only 17.6% of the time. The other 82% are 11–9, or 13–7, or worse.
Six strips of twenty flips. Three of them I really flipped. Three I made up by hand, trying to look random. Sort them.
- T H T T T T H T H T T H T T T T T H T H
- H T H T T H T H H T H T H T T H T H T H
- H T H H T H H H H T T H H H T H H T T H
- T H T H H T H T H T T H T H T H H T H T
- H H T T H T H H T T H T T H H T H H H H
- H T T H T H T H H T H T T H T H T H H T
The three real strips have longest runs of 5, 4 and 4. The three I invented have a longest run of 2 — every single one — and all three have exactly ten heads. I did not plan that. It is what a hand does when it is told to be random, and it took reading the table above to notice I had done it.
Why the brain gets it backwards
There is a reason, and it is not stupidity.
You are extremely good at spotting patterns. That is mostly a gift: it finds the path through the trees and the face in the crowd. But a run of five heads is a pattern, so when your brain generates a run of five it flags it as not random and quietly breaks it up.
So the picture in your head of "random" is really "no pattern anywhere", and that is a much rarer thing than randomness. Genuine randomness is lumpy. It clumps, it streaks, it leaves gaps. Smoothness is the fingerprint of something arranged.
The four things worth keeping
The book set out from a sheet of paper with too many sevens on it. Here is everything it turned out to need.
- Count the things that can happen, not the things you can see. Seven is a hill because six rolls make it. A room of twenty-three holds 253 pairs. This one move solved every puzzle here.
- Shares settle; counts do not. More rolling does not smooth the wobbles, it buries them under a bigger total.
- Nothing is ever due. A coin has no place to keep a debt, so a run does not make its opposite more likely. It only means an unlikely thing has already happened.
- Count the tries as carefully as the chance. One in a thousand is astonishing once and unremarkable across a hundred thousand goes.
A friend says a dice game is rigged because they rolled three sixes in a row. What is the most useful thing to ask?
- Whether the dice feel heavier on one side than the other
- How many rolls they have made altogether while watching for something odd
- Whether they were holding the dice in their left hand or their right
- How long ago the last three-sixes run was
That quiz answer is the whole book in one question. Three sixes in a row is 1 in 216. If your friend has been playing for an hour and rolling every few seconds, they have had hundreds of chances at it, so seeing one is expected rather than suspicious. If it happened on the first three rolls of their very first game, that is a different matter — but even then it is not proof of anything, because 1 in 216 things do happen, to somebody, roughly every 216 tries.
None of this makes chance safe or predictable. What it does is stop chance from being spooky. A run of luck is not a message, a rare event is not a sign, and nobody at the table is hot or cold. There are only ways things can come out, and how many of them there are.
That is countable, and now you can count it.