How to Size Bets You Can Survive: Bankroll Basics for Event Markets

Here is an uncomfortable truth about trading event markets: most people who blow up did not pick badly. They sized badly. They found a real edge, then bet so much on it that one ordinary loss — the kind that was always going to happen — took out months of gains. This guide is about the other half of the game, the half nobody posts screenshots of: how much to bet so that being occasionally wrong does not end you.

Why sizing matters more than picking

Start with the fact that any probability below 100% loses sometimes. A contract you correctly judge to be 90% likely will still lose one time in ten — not because you were wrong, but because that is what 90% means. Over a long series of good bets, losing streaks are not a possibility; they are a certainty. The question is never whether you will hit a rough patch. It is whether your account survives it.

That reframes everything. A great edge with reckless sizing is a slow-motion blow-up. A modest edge with disciplined sizing compounds for years. Between two traders with identical picks, the one who sizes to survive variance wins, and it is not close.

The ruin math, in plain numbers

Say you find bets where you have a genuine edge, and you stake a large fraction of your account on each. Consider what one bad run does.

If you put 25% of your bankroll on a single position and it loses, you are down to 75% — and now you need a 33% gain just to get back to even. String three such losses together (entirely normal, even with an edge) and you are at roughly 42% of where you started, needing to more than double to recover. The arithmetic of drawdowns is brutally asymmetric: losses compound against you faster than equal-sized gains dig you out.

This is risk of ruin — the probability that a string of losses drops you below the point of no return. It rises steeply with bet size. Halve your stake per position and you do not merely halve your risk of ruin; you push it down dramatically, because ruin depends on how deep a drawdown you can absorb before you are done.

The Kelly criterion: the theoretical ceiling

In 1956, John Kelly Jr. derived the mathematically optimal bet size for maximizing long-run growth when you know your edge. The Kelly criterion answers: given the odds and your estimated probability of winning, what fraction of your bankroll maximizes compounding over time?

The formula for a simple bet is:

**f\* = (bp − q) / b**

where b is the odds received on the wager, p is your probability of winning, and q = 1 − p. The result *f\** is the fraction of your bankroll to stake.

The key insight is not the algebra — it is what Kelly implies. Bet more when your edge is bigger and the odds are better; bet less when the edge is thin. And critically: Kelly defines a ceiling. Betting more than full Kelly does not just add risk for extra return — past that point, over-betting lowers your long-run growth while increasing your chance of ruin. It is strictly worse in both dimensions.

Why almost everyone should bet a fraction of Kelly

Here is the catch that professionals live by: full Kelly is optimal only if your probability estimate is exactly right. In the real world, your edge is an estimate, and estimates are noisy. If you think your edge is bigger than it truly is, full Kelly will over-bet you into deep drawdowns.

Because the growth curve near the Kelly peak is flat but the risk curve is not, betting half Kelly (or less) sacrifices only a little long-run growth while cutting volatility and drawdown substantially. Edward Thorp — who applied Kelly to blackjack and markets — long advocated fractional Kelly for exactly this reason. The rule of thumb most disciplined bettors converge on: estimate your Kelly fraction, then bet a fraction of that, precisely because you do not fully trust your own edge estimate. Humility, encoded as arithmetic.

The killer nobody sizes for: correlation

You can size every individual bet perfectly and still blow up if your bets are secretly the same bet.

Ten "independent" positions that all depend on the same macro outcome — one interest-rate path, one broad market direction, one shared catalyst — are not ten bets. They are one bet with ten times the size. When the shared driver goes against you, they lose together, and your carefully-sized book takes a coordinated hit that no single-position math anticipated.

This is the most underrated risk in event markets, because a correlated book feels diversified. It has many line items. But diversification is about independence of outcomes, not the number of tickets. Practical defenses:

• Map shared drivers. Before adding a position, ask what it has in common with what you already hold. If a single event would move several of them the same way, treat them as one larger position for sizing purposes.
• Limit exposure to any one theme. Whatever your per-bet size, keep total exposure to any single underlying driver well below your tolerance for a bad day.
• Prefer genuinely uncorrelated edges. A thin edge spread across truly independent outcomes compounds smoothly. The same edge stacked on one theme is a time bomb.

A survivable-sizing checklist

You do not need to be a quant to apply the ideas. A workable discipline:

1. Define your bankroll and never confuse it with your net worth. Only money you can genuinely afford to lose belongs in this account.
2. Estimate an edge honestly, then discount it. Assume you are a little overconfident, because you are.
3. Compute a Kelly fraction, then bet a fraction of it. Half or less is a sane default for uncertain edges.
4. Set a per-position maximum you never exceed, regardless of conviction. Conviction is exactly when people over-bet.
5. Track correlation across the whole book, not just per bet. Size the cluster, not the line item.
6. Decide your exit rule before the news, not during. Panic is not a sizing strategy.

Why this is the boring part that actually matters

None of this is glamorous, which is precisely why it is neglected — and precisely why it is where the durable results live. Picking is the part people enjoy; sizing is the part that decides whether you are still here in a year. If you internalize one thing: survive first, optimize second. A survivable edge, compounded patiently, beats a spectacular edge that ends in a crater.

Sizing pairs naturally with two other pieces on this site. To judge whether an "edge" is real before you size for it, read Calibration and Brier Scores. And for the specific structural tendency many event-market traders try to harvest — carefully, with exactly this kind of sizing discipline — see The Favorite-Longshot Bias.

Want to see disciplined sizing on a live, logged record? We are in an early testing period and giving free memberships to early testers — no card, no commitment. Join our Discord and DM the founder (or open a ticket) to claim: https://discord.gg/C6hX9w94Ej. The public dashboard shows every settled call, including the losses that sizing is designed to survive.

Independent research service. Not affiliated with Polymarket. Illustrative of past results, not a promise. Not investment advice.