Risk & portfolios

Choosing your risk: volatility targets, drawdown limits and position sizing

Risk & portfolios · Published 2026-10-25

A trading signal answers "what" and "which direction". It does not answer "how much". Yet the amount of risk taken often matters more to the outcome than the signal itself: a sound strategy run with too much leverage can be ruined by an ordinary bad month, and a good one run too timidly may not be worth its costs. This article reviews the main tools systematic investors use to choose their risk.

Start from the loss you can bear

Risk choices begin with constraints rather than formulas. What is the largest loss over a month or a year that would not force the strategy to stop, through a margin call, investor withdrawals or simple loss of nerve? A drawdown that seemed tolerable in a backtest can feel very different in real time. Writing down the maximum acceptable drawdown before trading gives every other tool a target.

Volatility targeting

Volatility targeting scales positions so that the portfolio's expected volatility stays near a chosen level, for example 10 per cent a year. If an asset's recent volatility is 20 per cent, a 10 per cent target implies holding half a unit of exposure; if volatility falls to 5 per cent, it implies two units, which requires leverage.

Volatility is estimated from recent returns, commonly with an exponentially weighted moving average that gives more weight to the latest observations. The appeal is that volatility is much more predictable than returns: high-volatility periods tend to follow high-volatility periods. Research on volatility-managed portfolios, including Moreira and Muir (2017) and, for momentum, Barroso and Santa-Clara (2015), found that scaling by recent volatility improved risk-adjusted returns in their samples. The approach reduces exposure after turbulence begins, but it cannot react to a sudden shock from a calm starting point.

Risk budgets across positions

Within a portfolio, volatility targeting can be applied at each level. Inverse-volatility weighting gives each position similar standalone risk. Risk parity goes further, equalising each position's contribution to total portfolio risk after accounting for correlations. These methods ensure that a single volatile asset does not dominate, but they say nothing about which positions have the best expected return. Our article on combining diversified alpha discusses weighting between strategies.

Fixed-fractional sizing

A widely used rule in trading literature is to risk a fixed fraction of capital on each trade, for example one per cent, where the risk is the distance to the stop-loss. If the stop is 5 per cent below the entry, a one per cent risk implies a position worth 20 per cent of capital. Fixed-fractional sizing automatically shrinks positions after losses and grows them after gains, which limits the chance of ruin from a losing streak.

The Kelly criterion

John Kelly (1956), working on information theory at Bell Labs, derived the bet size that maximises the long-run growth rate of capital. For a simple bet with win probability p and net odds b, the Kelly fraction is f* = (bp − (1 − p)) / b. For a continuous investment with expected excess return μ and variance σ², the corresponding leverage is approximately μ / σ². Edward Thorp later applied the criterion to blackjack and to investing.

Kelly sizing has a strong theoretical justification, but full Kelly is aggressive: it produces very large drawdowns, and it is extremely sensitive to the estimate of μ, which is the least reliable input. Overestimating the edge leads to over-betting, and betting more than twice the Kelly fraction produces negative long-run growth. Many practitioners therefore use a fraction of Kelly, such as half, which gives up a modest share of the growth rate for a large reduction in volatility.

Drawdown controls

Some strategies reduce exposure as losses accumulate: for example, halving positions after a 10 per cent drawdown and stopping after 20 per cent. Grossman and Zhou (1993) studied the optimal policy under a constraint that wealth never falls below a fixed fraction of its previous peak. Drawdown rules protect capital but can lock in losses just before a recovery. They work best as circuit breakers that trigger a review rather than as mechanical switches.

Leverage and its hidden costs

Leverage amplifies both returns and losses, but it also adds risks that do not appear in a backtest: financing costs, margin calls at the worst moment, and the possibility that a lender changes terms. Our article on prime brokers explains how margin works for hedge funds. A useful test is to ask what happens if volatility doubles overnight and the broker raises margin requirements at the same time.

Stop-losses

A stop-loss closes a position when its loss reaches a set level. Stops limit the damage from a single trade, but their effect on a strategy's overall return depends on the return process. Kaminski and Lo (2014) showed that stop-loss rules can add value when returns show momentum, and tend to reduce returns when prices follow a random walk. Like everything else, stop levels should be tested out of sample rather than tuned to the past.

Measuring the risk you chose

After choosing a sizing method, measure what it implies: expected volatility, value at risk and expected shortfall, maximum historical drawdown under the sizing rule, and the leverage required at its peak. Running the sizing rule through stressed historical periods, such as 1987, 2008 or March 2020, shows whether the chosen risk is compatible with the loss written down at the start.

Summary

• Decide the maximum tolerable loss first; let it constrain everything else.
• Volatility targeting is well supported but reactive.
• Kelly sizing is theoretically optimal for growth and dangerously sensitive to estimated edges; fractional Kelly is the usual compromise.
• Drawdown rules and stops limit damage at the cost of sometimes cutting too late or too early.
• Leverage adds financing and liquidity risks that backtests omit.