Most traders obsess over what to buy and barely think about how much. That is backwards. You can pick the right direction more often than not and still go broke by sizing badly — too large into a drawdown, too timid into your best ideas. The Kelly criterion is the math that fixes this.
What the Kelly criterion actually says
The Kelly criterion, published by John Kelly in 1956, answers one question: what fraction of your capital should you risk on a bet to maximise long-run growth? Bet too much and a losing streak wipes you out; bet too little and your capital compounds slower than it should. Kelly finds the growth-optimal middle.
For a simple bet, the formula is f = (bp − q) / b, where p is your probability of winning, q = 1 − p is your probability of losing, and b is the payoff ratio (how much you win per unit risked). The result f is the fraction of your bankroll to put on the trade.
A worked example
Suppose your edge says a setup wins 55% of the time (p = 0.55, q = 0.45), and when it wins you make 1.5x what you risk (b = 1.5). Plug in: f = (1.5 × 0.55 − 0.45) / 1.5 = (0.825 − 0.45) / 1.5 = 0.375 / 1.5 = 0.25. Kelly says risk 25% of your bankroll on that trade.
Twenty-five percent on a single trade feels enormous — and that reaction is the most important lesson in this article.
Why nobody trades full Kelly
Full Kelly is growth-optimal in theory but brutal in practice. Its drawdowns are larger than almost any human can stomach, and it assumes you know p and b exactly — you do not. Your edge is an estimate, and overestimating it makes Kelly dangerously aggressive.
The standard fix is fractional Kelly: bet a fraction of the Kelly number — commonly half. Half-Kelly captures roughly three-quarters of the growth with far smaller swings, and it cushions the damage when your edge estimate is too optimistic. In the example above, half-Kelly would risk about 12.5% rather than 25%, and many traders go lower still.
How to use it without a PhD
- Estimate your edge honestly from real, out-of-sample results — not from your best week.
- Apply a fraction of Kelly (half or less) to leave room for error in that estimate.
- Cap any single position regardless of what the formula says, so one bad read is never fatal.
- Re-estimate as your win rate and payoff ratio drift; Kelly is only as good as its inputs.
The takeaway is not the formula itself — it is the mindset. Size positions in proportion to your edge, not your conviction or your excitement. The trader who survives a losing streak is the one who sized for it in advance.
This is how Sintinel approaches sizing: a stronger, better-corroborated composite signal earns a larger suggested allocation, with hard caps that keep any one ticker from dominating the portfolio. The math does the sizing so conviction does not.