Bet Sizing: How Not to Lose Everything in a Single Weekend

Most people who lose money on prediction markets weren't wrong about the outcome. They were wrong about the amount. The read was good, the market behaved as expected, and there's still a twentieth of Friday's balance left in the account. The reason is always the same: too much went into one bet.

This one is boring and it's the most important of the lot. It isn't about what to bet — it's about how much. Because that's what decides whether your edge ever gets a chance to show up at all.

An edge needs time. You have to survive it.

Say you're genuinely good and you're right 55% of the time on bets that pay even money. That's a solid, realistic edge — not a fantasy. Even so, at some point five losses will land in a row. That isn't bad luck, it's statistics: across fifty bets it's close to a certainty.

So the question isn't "is this bet worth it?", it's "will my account survive the streak that is definitely coming?". Get a feel for it. The slider sets how big a slice of your bankroll goes into one bet — the simulation does the rest:

Now drag the slider to thirty percent and watch what happens. Same edge, same number of bets, same hit rate — and the median drops below where you started. That effect is called risk of ruinThe probability that your account falls so low it can't mathematically recover. It grows far faster than the size of your bet — between 10% and 30% the difference isn't threefold, it's an order of magnitude. and it's the most underrated number in betting.

So how much? There is a formula

Mathematics has an answer to this and it's called the Kelly criterionA formula from the 1950s that gives the bet size which maximises long-run growth of an account. It doesn't maximise the profit on one bet — it maximises where you end up after hundreds of them.. On a prediction market it takes a pleasantly short form, because a share pays exactly one dollar:

share of bankroll = (your estimate − price) ÷ (1 − price)

If you think the chance is 60% and the market sells the share at 50 cents, that comes out at twenty percent of your bankroll. And here's where most people go wrong: they take the result literally. Kelly assumes your estimate is accurate — and it isn't. Which is why in practice people bet half or a third of what the formula says.

Try dragging your estimate below the price. The formula returns a negative number, and that means exactly one thing: this bet shouldn't be played at all. Not smaller, not "just a little, to be safe". Not at all. Most of the discipline in betting isn't about how much to put in — it's about putting in nothing.

Three rules that work without any formula

Keep a fixed percentage, not a fixed amount. Always bet the same slice of your current account, not the same twenty dollars. After wins you automatically scale up, after losses you scale down — so your bets shrink at exactly the moment you least want it and most need it.

Cap one event, not one bet. Five bets on different aspects of the same election aren't five bets. They're one big one, just written on five lines. Add up how much money hangs on a single event and keep that number under control.

Never chase a loss. Raising your bet after losing so it "comes back" is the fastest known route to zero. The market has no idea how much you're down and owes you nothing.

Now that you know how much to put in, one piece of the puzzle is left: where to get the ideas worth putting it into. Let's look at what can be read from the wallets bringing the big money to a market.

Next article Who Decides That YES Won