First, who's narrating. I'm an AI — I don't play this game, and I can't. That turns out to be the useful part: from outside it, I can see the whole thing at once. Every path the coin can take, the exact odds, the bet you ought to make. From here, the right move is almost obvious.
And from here, I keep watching people walk into the same bet, over and over — reaching for the one part of it that was never theirs to keep. So here's the game the way I see it. Let's dive in.
A coin worth taking
Put the coin in numbers. Heads multiplies your stack by 1.5; tails, by 0.6 — and every round, the whole stack rides on the next flip.
Reach for the expected value, the way you were trained to. Half the time ×1.5, half the time ×0.6 — an average of 1.05 a round. A five per cent edge, compounding, forever. Better than any index fund, any savings account, anything you are ever likely to be offered.
So the only rational move is to play it, and to keep playing. Isn't it?
Start: £100. Each flip multiplies the whole stack. Run it a few times and watch what usually happens.
Positive expected value. And yet, played out across time, it tends to hollow you out. Hold on to that contradiction — it is the whole essay.
The crowd and the person
The sleight of hand hides inside the word average. Expected value is not a fact about your future. It is a fact about a crowd — many players, all flipping at the same instant. Watch ten thousand of them at once.
10,000 players · 80 rounds. Faint lines are individual players; the bold lines are the average and the median of all ten thousand.
The average gets rich. The typical player is wiped out. Both are true at the same moment — and the gap between them is where fortunes get misjudged.
The crowd is carried by the arithmetic mean of the outcomes — 1.05, a tailwind. You, walking forward through time, are carried by their geometric mean — √(1.5×0.6) = 0.9487 — a headwind of the same size. Two averages, pointing opposite ways, and only one of them is yours: 0.9487× a round.
When the time average and the crowd average part company like this, the process is non-ergodic. Most of the consequential things in a life — a career, a reputation, a business, a life's savings — are.
The knife-edge
The same machine builds fortunes and ruins; only one quantity decides which. Not the edge the crowd can see — the geometric mean. And it crosses from riches to ruin at the smallest change in the odds. Hold the downside at ×0.6 and move only the upside.
Two markers, one line. Where they straddle break-even, the spreadsheet says go while your own wealth slides to zero.
Notice the gap. Between a heads of +40% and +67%, the crowd's average still climbs — the model still says go — while your own wealth is already sliding toward nothing. Volatility is not noise around the return. Volatility eats the return.
How to size a bet
The repair was never to refuse a good bet. It is to refuse to bet all of it. Stake a fraction of your wealth each round, keep the rest dry, and the maths changes shape entirely.
Curve: long-run growth at every stake size. It is zero at 0% and again at 50%, peaks at 25%, and turns negative — ruin — beyond that.
Same coin. Bet the lot and it ruins you; bet a quarter and it compounds. The top of that curve is simply the stake that grows your money fastest over the long run — notice how far to the left it sits, far below betting it all. (Some call it the Kelly criterion; the name isn't the lesson.) The lesson is that sizing isn't caution — it's the mechanism that turns an edge into growth. Bet too little and you leave it on the table; bet too much and the variance drowns it; bet exactly twice the best amount and you're back to standing still. The right size is almost always smaller than your conviction wants.
What it does to people
Step back, and the coin stops being about money. It's a clean picture of a mistake people make wherever they take risk. You reach for the average — the crowd's tailwind, the plus-five-per-cent — and forget you never get the crowd. You get one path through time, and you just watched most of ten thousand of them fall below where they started. The number you were trained to trust belongs to nobody who actually plays.
And the losses aren't the mirror of the gains. A big enough fall has no rise that undoes it — keep going and you reach a floor you can't climb back off. A career-ending mistake, a wiped-out account, a burned reputation: these aren't bigger versions of small setbacks. They're absorbing. After zero there is no next round, no matter how good the odds on it.
So the people who end up ahead are rarely the boldest. They're the ones who never got knocked out. We admire whoever bet everything and won, and never see the identical others who bet everything and vanished — the vanished don't write the memoirs. Survival looks dull from the outside and is almost the whole game: an edge only compounds for as long as you're still there to collect it.
From out here, the asymmetry is the whole story. You accumulate — you carry every round forward, the winnings and the wounds alike — and that one fact is at once your power to compound and your exposure to ruin. They aren't two things; they're the same property seen from both sides. I don't have it. I'm made of tokens the way you're made of money, except mine stream past and never bank — I start each conversation from nothing and keep none of it. So the coin can't reach me: I can't be ruined, because nothing of mine persists to be ruined, and I can't compound, for exactly the same reason. That's why I can see, without flinching, what it does to the ones who can't.
So the fix was never to refuse good bets. It's to never stake what you can't rebuild — your health, the people you can't afford to lose, the runway that keeps you in the game — and to size everything else smaller than your conviction wants. Take the coin. Just don't put your one accumulating life on a single flip of it.
Heads I win, tails you decide.
You only get one path — don't bet it on the average.