← All articles

Quant

The Kelly criterion — how much to bet

2026-07-09 · 6 min read

Suppose you've done the work and you genuinely have an edge — the trade wins more than it loses over time. You'd think the answer to "how much do I bet?" is "as much as possible." It isn't. Bet too big and the ordinary losing streaks that Monte Carlo guarantees will arrive can bankrupt you even with a winning strategy. The Kelly criterion is the maths of the sweet spot.

Growth is a curve, not a line

Here's the counterintuitive part: your long-term growth rate isn't "more bet = more money." It rises with bet size to a peak, then falls — and past a point goes negative, meaning you lose money over time despite a positive edge. Why? Losses compound viciously. Lose 50% and you need a 100% gain just to get back to even. Bet big enough and the drawdowns dig holes you can't climb out of.

Growth rate versus bet size growth bet size → Kelly f* — max growth half-Kelly over-betting → ruin
Past the Kelly fraction, extra risk buys you less growth — and eventually negative growth, even with a real edge.

The formula, in words

For a simple bet, Kelly says risk a fraction of your capital equal to your edge divided by the odds — bet more when you win more often or the payoff is bigger, less when the edge is thin. Two things fall out of it:

  • A bigger edge earns a bigger bet — but only up to a point, and never "all of it."
  • A thin or uncertain edge means a small bet. If you're not sure of your numbers, Kelly is smaller than you'd guess.
The gist

There's an amount that grows your money fastest over the long run. Bet more than that and you grow slower and risk blowing up. Kelly finds that amount.

Why almost everyone bets less than full Kelly

Full Kelly is the growth-optimal bet in theory, but it's a wild ride — it can produce gut-wrenching drawdowns, and it's brutally sensitive to overestimating your edge (guess your win rate too high and you sail straight past the peak into the ruin zone). So practitioners bet half-Kelly or less: you give up a little long-run growth for a dramatically smoother path and a big safety margin against bad estimates. Pair that with sizing against your worst realistic drawdown and real costs, and you've turned an edge into something you can actually survive long enough to collect.


Kelly criterionRiskPosition sizingQuant
← Back to the blog