What it means
The Kelly criterion is a position-sizing formula that maximizes geometric (compound) growth of capital. Formula: f* = (bp - q) / b, where b = ratio of win to loss size, p = win probability, q = loss probability. f* is the optimal fraction of capital to risk per trade. For a system with 60% win rate at 1:2 RRR: f* = (2×0.6 - 0.4) / 2 = 0.4 = 40% of capital per trade — too aggressive in practice.
Why it matters
Kelly proves there's an OPTIMAL position size, not just 'as much as the broker allows.' Sizing above Kelly produces volatility destruction (large drawdowns destroy compound growth faster than the gains can recover). Sizing below Kelly leaves return on the table. Real-world practice: use 'fractional Kelly' (typically 0.25 × Kelly or 0.5 × Kelly) to balance growth and survivability.
How to use it
Compute Kelly from your system's measured stats. Apply fractional Kelly (commonly 0.25 of full Kelly) for actual position sizing. The fractional approach accepts ~50% of full-Kelly growth in exchange for ~75% lower drawdown — usually the right tradeoff for retail traders.
System with 55% win rate, average win +1.5R, average loss -1R. Kelly: f* = (1.5×0.55 - 0.45) / 1.5 = 0.25 = 25% of capital per trade. Far too aggressive (would tolerate massive drawdowns). Quarter-Kelly: 6.25% per trade — still aggressive. Most retail traders cap at 1-2% per trade, far below Kelly, accepting slower growth for psychological survivability.
Why full Kelly is psychologically impossible to trade
Full Kelly maximizes geometric growth IN EXPECTATION but produces extreme volatility. A 25% Kelly system can have 50-80% drawdowns en route to compound returns. Psychologically, very few traders can hold position through 50%+ drawdown without intervening — and any intervention destroys the math. Half-Kelly cuts drawdown to ~30-50% while still producing ~75% of full-Kelly growth. Quarter-Kelly produces ~50% of full-Kelly growth with 20-30% max drawdown — survivable for most traders.
Kelly assumes stable edge and independent trades
Kelly's math requires: (1) accurate win-rate and RRR measurements (which are themselves uncertain), (2) trade independence (no correlation between consecutive trades), (3) stable edge over time. None of these hold perfectly in trading. Underestimating any of these means real-world Kelly is HIGHER than computed Kelly — another reason to use fractional Kelly as a safety margin.
Frequently asked
Should I actually use Kelly to size my trades?
Use fractional Kelly (0.25× to 0.5×) as a UPPER BOUND on position size. Most retail traders should size below quarter-Kelly because (a) edge estimates are noisy, (b) drawdown tolerance is low. Kelly is a ceiling, not a target.
Why does over-sizing destroy compound returns?
Compound math: a 50% drawdown requires a 100% recovery to break even. A 75% drawdown requires 300% recovery. Above Kelly sizing produces drawdowns that take asymmetric returns to recover. Below Kelly sizing produces smoother equity curves with fewer recovery cycles.
Does Kelly account for correlated positions?
No — Kelly assumes independent trades. For correlated portfolios (multiple FX longs all betting on USD weakness, for example), the EFFECTIVE Kelly is much lower than the per-position Kelly. Treat correlated positions as a single position for Kelly sizing.
How is Kelly different from fixed-fractional sizing?
Fixed-fractional risks the same % per trade (e.g., always 1% of equity). Kelly varies position size based on edge (higher Kelly = bigger size). Most practical sizing is hybrid: fractional-Kelly capped at a fixed maximum like 1-2% per trade.
Want a worked example or a deeper dive? Ask Rocky how this concept applies to your specific watchlist or trade idea.
Ask Rocky