Kelly criterion explained: optimal bet sizing for traders

By Ken Chigbo, Founder, KenMacro. Published 2026-05-12.

What is Kelly criterion?

Kelly criterion is the optimal bet-sizing formula derived by John L Kelly Jr at Bell Labs in 1956. The formula gives the fraction of capital to risk per trade that maximises the long-run geometric growth rate of capital, given the strategy's win rate (w) and average win-to-loss ratio (b). The simplified form for a binary win-loss strategy is f equals w minus (1 minus w) divided by b, where f is the optimal fraction of capital to risk. A strategy with a 60 per cent win rate and a 2:1 average win-to-loss ratio gives Kelly equals 0.60 minus 0.40 divided by 2, equal to 0.40, or 40 per cent of capital per trade. The figure is mathematically optimal under perfect inputs but is extremely volatile in practice.

How traders use Kelly criterion

Professional traders rarely run full Kelly because the formula assumes the win rate and win-to-loss ratio are perfectly known, which they never are. Estimation error in the inputs amplifies in the output: a small overestimate in win rate produces a large overestimate in Kelly fraction, leading to catastrophic drawdowns. The practical version is half-Kelly (50 per cent of the Kelly fraction) or quarter-Kelly. Half-Kelly gives roughly 75 per cent of the long-run growth rate of full Kelly with materially lower drawdowns. Quarter-Kelly is even more conservative and is the level the desk recommends for new traders. The desk's deployed frameworks size positions at 0.5 to 1.5 per cent of capital per trade, which corresponds to roughly quarter-Kelly to eighth-Kelly for typical positive-expectancy strategies. Anyone running full Kelly on a real account is taking a level of risk that almost guarantees a 50 to 80 per cent drawdown over a multi-year horizon.

Worked example with Kelly criterion

Consider a swing strategy with a 55 per cent win rate, an average win of 2R, and an average loss of 1R. The win-to-loss ratio b is 2. Full Kelly equals 0.55 minus 0.45 divided by 2, equal to 0.325, or 32.5 per cent of capital per trade. On a 10,000 dollar account, full Kelly recommends risking 3,250 dollars per trade. The volatility of capital at that sizing is extreme: drawdowns of 60 to 80 per cent are routine within 50 trades, even with positive expectancy. Half-Kelly cuts the recommendation to 16.25 per cent (1,625 dollars per trade), still aggressive. Quarter-Kelly cuts it to 8.13 per cent (813 dollars per trade), and the desk's deployed sizing of 1 to 1.5 per cent (100 to 150 dollars per trade) is roughly one-twentieth Kelly, prioritising drawdown control over absolute long-run growth.

Join the Macro Mastery desk

Frequently asked

What is the Kelly criterion formula?

The simplified Kelly formula for a binary win-loss strategy is f equals w minus (1 minus w) divided by b, where f is the optimal fraction of capital to risk, w is the win rate, and b is the average win-to-loss ratio. A 60 per cent win rate and 2:1 win-to-loss ratio gives Kelly equals 0.40, or 40 per cent of capital per trade.

Why is full Kelly too aggressive?

Full Kelly assumes the win rate and win-to-loss ratio are perfectly known. In practice they are estimated from historical samples and carry uncertainty. Small estimation errors in the inputs produce large errors in the Kelly fraction. Drawdowns of 60 to 80 per cent are routine at full Kelly even with positive-expectancy strategies. Half-Kelly or quarter-Kelly is the practical version.

What position size does the KenMacro desk recommend?

The KenMacro desk's deployed frameworks size positions at 0.5 to 1.5 per cent of capital per trade, which corresponds to roughly quarter-Kelly to eighth-Kelly for typical positive-expectancy strategies. The conservative sizing prioritises drawdown control and survival over absolute long-run growth. New traders typically benefit from staying at 0.5 to 1 per cent per trade until the strategy is validated.

Educational analysis only. Past performance does not guarantee future results. Manage risk against your own portfolio.