Guides / Kelly Criterion Position Sizing, Explained

f = p − q/b

Kelly criterion position sizing, explained with real numbers.

Kelly tells you the position size that maximizes long-run growth for a known edge. It also explains why almost nobody should run full Kelly. Here's the formula, what it actually optimizes for, and what half-Kelly buys you.

Cephic TeamUpdated Nov 2, 20267 min read

Quick answer: the Kelly criterion is f = p − q/b, where p is your win rate, q is 1 − p, and b is your reward-to-risk ratio. The result is the fraction of your account to risk per trade that maximizes long-run compound growth, not the fraction that minimizes drawdown. Most traders who use Kelly at all use a fraction of it, commonly half-Kelly, trading some growth rate for a meaningfully smaller chance of a severe drawdown along the way.

THE FORMULA

What Kelly is actually solving for

The Kelly criterion answers one specific question: given a win probability and a payout ratio, what fraction of the account, risked on every trade, grows the account fastest over many trades? For a binary win/loss bet with a fixed reward-to-risk ratio, the formula is:

Kelly % = (p − q / b) × 100, where p = win rate, q = 1 − p, b = reward-to-risk ratio

This is the exact formula Cephic's Risk of Ruin calculator uses to compute the Kelly and half-Kelly figures it shows next to your simulation results. If the result works out to zero or negative, there's no positive edge at that reward-to-risk, and the formula returns 0% rather than a negative size. A negative Kelly fraction means "don't take this bet," not "bet against it."

A WORKED EXAMPLE

Running the numbers on a 55% win rate, 1.2 reward-to-risk strategy

p = 0.55, q = 0.45, b = 1.2. Kelly % = 0.55 − (0.45 / 1.2) = 0.55 − 0.375 = 0.175, or 17.5%. Full Kelly here says risk 17.5% of the account on every trade. Half-Kelly, the more commonly used fraction, is 8.75%. Both numbers are a long way from the 1-2% most traders default to, which isn't a sign that 1-2% is wrong. It's a sign that full Kelly is sized for a very different objective than most traders actually have.

WHY FULL KELLY IS RARELY RUN LIVE

Full Kelly optimizes growth, not comfort

Kelly-sized betting maximizes the expected long-run growth rate of the account. It does not minimize volatility, drawdown, or the chance of a rough multi-month stretch, the growth-maximizing size is also, by construction, a size that produces large swings along the way. Betting at full Kelly with a real edge still produces drawdowns of 30-50% with real frequency over enough trades, because the formula has no term for how much pain a trader can tolerate before abandoning the strategy, or the account.

There's a second problem, and it's the bigger one in practice: Kelly needs an exact win rate and reward-to-risk as inputs. Live trading doesn't supply exact numbers. It supplies a backtest-estimated win rate that drifts, and a reward-to-risk that depends on where trades actually get filled and stopped out, not where they were planned on a chart. Feed Kelly an edge that's a little optimistic (which is the normal state of a backtested edge the first few months live) and the "optimal" size it returns is already too large before a single live trade happens.

WHAT HALF-KELLY TRADES AWAY

Half-Kelly: most of the growth, a lot less of the swing

Halving the Kelly fraction is a common compromise because of how the growth-vs-drawdown trade-off bends near the peak: cutting position size in half from full Kelly gives up roughly a quarter of the theoretical growth rate, but cuts the typical drawdown by much more than half. It's a well-known property of the Kelly growth curve: the curve is flat near its peak and steep on the way down, so moving a bit away from the peak costs little in growth and buys a lot of room against a bad stretch. That asymmetry is the entire argument for half-Kelly, or quarter-Kelly for traders who want even more room.

It's still a position-sizing ceiling built on an assumed edge, not a guarantee of anything. If the underlying win rate or reward-to-risk is wrong, half of a wrong number is still a wrong number, just a smaller one.

WHAT TO DO WITH THIS

Use Kelly as a reference point, not a position-sizing instruction

Run your win rate and reward-to-risk through Cephic's Risk of Ruin calculator and look at the Kelly and half-Kelly figures it returns alongside the actual ruin probabilities at your chosen risk per trade. The ruin probabilities matter more than the Kelly number by itself, because they show what your sizing does to your specific account over your specific trade count, not just a theoretical growth-maximizing fraction based on an edge that might drift. If your planned risk per trade sits well under half-Kelly, the Kelly number is telling you that you have room, not that you should use all of it.

Check where your risk per trade sits relative to your own Kelly fraction.

Open the calculator

Related free calculators: Risk of Ruin Calculator (shows Kelly) · Position Size Calculator