Investors obsess over what to buy. The question that more often decides survival is how much. That's position sizing: the fraction of your capital committed to any single idea.
Suppose you found a genuinely favorable bet — positive expected value, meaning the average outcome across many repetitions is a profit. Bet too little and the edge barely moves your wealth. Bet too much and you can still go broke, because excessive size produces violent swings. Lose 50% and you need a 100% gain just to get even. Lose 90% and you need 900%. That asymmetry is the deepest fact in risk management.
For any strategy with an edge and risk, there's a size that grows wealth fastest, sizes below it that grow slower but calmer, and sizes above it that mathematically guarantee worse outcomes. Remarkably, there's a formula for the fastest-growing size. A Bell Labs physicist named John Kelly published it in 1956.
The Kelly criterion answers one precise question: what fraction of your bankroll — your total risk capital — should you stake on a favorable bet, repeated many times, to maximize the long-run growth rate of your wealth?
For a simple bet: Kelly fraction equals p minus q divided by b. Here p is your probability of winning, q is the probability of losing, and b is the net odds — how many dollars you win per dollar staked when you win. Win $1 per $1 staked and b is 1; win $2 per $1 and b is 2.
Two properties matter. If the formula outputs zero or a negative number, you have no edge — the correct stake is nothing, and no sizing scheme rescues a losing bet. And the output is a fraction of current bankroll, recalculated as wealth changes: you bet fewer dollars after losses and more after wins, which makes total ruin impossible under strict Kelly. Drawdowns — declines from a peak — remain very possible, as we're about to see.
Take a coin that lands heads 60% of the time, paying even money: stake $1, win $1 on heads, lose the stake on tails. No market offers anything close, which makes it a clean laboratory.
Plug in: p is 0.6, q is 0.4, b is 1. Kelly fraction equals 0.6 minus 0.4 divided by 1, which is 0.20. Bet 20% of your bankroll each flip. On $10,000, the first bet is $2,000; win and the next is 20% of $12,000, or $2,400; lose and it's 20% of $8,000, or $1,600.
At that size, wealth compounds at about 2% per flip on average — the maximum achievable for this bet. Now overbet. At 40% per flip, twice Kelly, the average growth rate falls to roughly zero: a 60/40 coin in your favor, and you merely tread water while riding stomach-turning swings. Push past twice Kelly and expected growth turns negative — you'll go broke with mathematical certainty over time. That asymmetry again: a win then a loss at 40% turns $100 into $140, then $84. Volatility itself taxes compound growth, and oversizing feeds the tax. Above the Kelly point, more aggression means less money.
So just bet full Kelly? Almost no professional does, for two hard reasons.
First, the ride. Full Kelly maximizes growth by accepting brutal volatility. A standard result: betting full Kelly, the chance your bankroll gets cut in half at some point is about one in two, and the chance it drops to a tenth is about one in ten. Most humans abandon a strategy long before a 50% drawdown resolves, and abandoning at the bottom converts a temporary drawdown into a permanent loss.
Second, and worse: the formula's inputs are lies. In the casino you know p exactly. In markets you estimate your edge from historical data, and those estimates are noisy, biased upward by every flattering choice in your backtest, and prone to decay as others find the same trade. The math is asymmetric too — overbetting cuts growth faster than underbetting by the same amount — so an optimistic edge estimate means overbetting the truth, the one error the formula punishes without mercy. Uncertainty about your edge systematically pushes the correct bet smaller.
The standard professional answer is fractional Kelly: compute the full Kelly stake, then bet a fixed fraction of it — most commonly half.
Near the optimum, the growth-versus-size curve is flat on the low side and steep on the high side. Half Kelly delivers about 75% of the maximum growth rate with roughly half the volatility, and the chance of ever halving your bankroll falls from about one in two to about one in eight. You give up a quarter of the growth to escape most of the pain — and gain a safety margin, since half of an inflated Kelly estimate may land near the true optimum anyway. Quarter Kelly is defensible for strategies whose edge you trust least.
For investments rather than bets, practitioners use a continuous version: the Kelly fraction is roughly the expected excess return divided by the variance of returns, where variance is volatility squared. Plug in market-like numbers — say a 5% excess return and 16% volatility — and you get roughly 2, meaning full Kelly would hold about twice your wealth in stocks, using leverage. Ed Thorp, the blackjack and hedge fund pioneer who popularized Kelly, spent decades warning that estimated inputs make full Kelly a trap. The formula is only as good as the guesses you feed it.
You may never compute a Kelly fraction. The framework still hands you four durable lessons.
No edge, no bet. If you can't articulate why an idea has positive expected value — and why the person on the other side is wrong — the correct size is zero. Sizing discipline starts with the honesty to bet nothing.
Halve your confidence. Whatever edge your analysis or backtest suggests, the live version will almost certainly be smaller: markets adapt, and backtests flatter. Sizing as if you have half your measured edge is the cheapest insurance in quantitative investing.
Diversification changes the math. Kelly applies to correlated ideas jointly, not one at a time. Ten positions that crash together — ten tech stocks, say — are closer to one big bet than ten small ones, and sizing them independently is how portfolios drift past full Kelly overall.
Respect the drawdown you can't survive. The right size isn't the one that maximizes theoretical growth. It's the largest size whose worst plausible stretch you'll actually hold through — financially and psychologically. Growth formulas assume you keep playing. Making sure you can is the real job.
Previous: Pairs Trading and Statistical Arbitrage · Next: Risk Parity and Portfolio Weighting Schemes · Financial glossary
← Back to all investing concepts