Kelly criterion in gold trading

The Kelly criterion calculates the fraction of capital to risk on each bet to maximize the long-run growth rate, given the probability of winning and the payoff. For trading it is often written as f = W − (1 − W) ÷ R, where W is the win rate and R is the average win divided by the average loss.

How it applies to gold trading

Kelly assumes you know your true win rate and payoff exactly and that they never change. A trader’s real numbers come from a limited sample and drift with market conditions, so full Kelly almost always overestimates the safe risk. Full Kelly also produces very large drawdowns even when the inputs are correct; halving the bet size gives up only a little growth while cutting volatility sharply. For these reasons professionals who use Kelly at all apply a small fraction of it, and treat it as an upper limit rather than a target. This entry explains the idea; it is not a recommendation for any risk level.

Worked example

Example (hypothetical numbers): a gold strategy wins 45% of trades with an average win twice the average loss. f = 0.45 − 0.55 ÷ 2 = 0.175, or 17.5% of the account per trade. On a $10,000 account that is $1,750 at risk per trade, a level at which five losses in a row, which this strategy will see regularly, would cut the account by more than 60%. A quarter Kelly would be about 4.4%, still above the 0.5% to 2% many traders use.

A common mistake

Plugging a small backtest’s statistics into the formula and trading the full result. Overestimate the edge slightly and full Kelly turns growth into ruin.

See also: Position sizing, Risk of ruin, Risk per trade, Expectancy.

Frequently asked questions