Volatility targeting:
the rule and the reality.

What is volatility targeting?

Volatility targeting adjusts the size of an investment exposure toward a chosen level of return variability. In a simple one-asset model, an asset estimated to have 20% annual volatility needs 50% exposure to produce 10% modeled portfolio volatility, with the rest held in constant-value cash.

The target is a scale for sizing. It does not tell you whether the asset has positive expected returns, whether it fits your circumstances, or how much you could lose. Realized volatility may differ substantially from the estimate.

Both volatility inputs must use the same time scale and units. A daily estimate cannot be compared directly with an annual target. Annualizing a daily estimate commonly uses a square-root-of-time assumption, but that assumption can be unreliable when returns are dependent or volatility changes.

One target. Three very different positions.

Assume capital of 100,000, a 10% annual volatility target and a 1× exposure cap. Cash earns zero; there are no costs, taxes or market moves during resizing.

When the estimate doubles from 20% to 40%, the rule halves exposure. When the estimate falls enough, the cap prevents the target from being reached. Neither outcome says what the next return will be.

The estimate is part of the strategy.

Historical volatility summarizes past returns. It is not a measurement of tomorrow’s risk. A short lookback can respond quickly to new observations but can also produce unstable sizing. A long lookback can be steadier while reacting slowly to a sudden change.

Published methodologies vary. For example, FTSE Russell’s Custom Volatility Target Index rules combine short- and long-horizon exponentially weighted estimates, a lag and a maximum exposure. Those choices are substantive parts of that index, not implementation details interchangeable with any other rule.

Our calculator takes an estimate you supply. It does not choose a lookback, download market data or claim to reproduce that index.

What can go wrong?

Volatility arrives before the estimate catches up.

If exposure was sized on a calm period, a sudden price move can occur before the rule responds. Increasing the estimate afterward does not reverse the loss. Reducing exposure after a fall can also leave less exposure during a rebound.

Low estimated volatility can request leverage.

A 10% target divided by a 5% estimate produces 2× exposure. Borrowing, derivatives or other leveraged implementations bring financing, collateral and liquidation constraints. The SEC’s leverage bulletin describes how borrowing can magnify losses and expose an investor to losses beyond invested capital. A mathematical exposure is not evidence that it is available, affordable or suitable.

Rebalancing is not free.

Changing the estimate can mean changing the position repeatedly. Spreads, commissions, market impact, taxes and financing can change the result. Trading less frequently or adding a rebalance threshold changes both costs and how closely exposure follows the model. A comparison should specify the same execution assumptions for every strategy.

A cap on exposure is not a cap on loss.

An exposure cap limits the size requested by the rule. It does not limit the asset’s move, eliminate gaps or guarantee liquidity. A 10% volatility target is not a promise that a portfolio cannot lose more than 10%.

Several holdings introduce correlations.

For a portfolio, combine the weights, individual volatilities and correlations before inferring risk. Summing the volatility targets of individual positions is not a portfolio volatility calculation.

For two equal-weight assets, each with 20% volatility, the model gives 20% portfolio volatility at correlation +1, about 14.14% at correlation 0, and 0% at correlation −1. The last case is an exact mathematical cancellation under the assumptions, not a practical risk-free allocation. Estimates change, correlations can shift and the simple model leaves out other risks.

Volatility sizing, stop-loss sizing and Kelly.

Volatility sizing scales exposure using a variability estimate. Stop-loss sizing uses the distance between an entry and a proposed stop to translate a loss budget into units; actual execution may occur at a worse price. Kelly sizing needs probabilities and payoffs and optimizes expected logarithmic growth under its model.

These methods answer different questions. A volatility target alone supplies neither a stop price nor an expected-return edge. Try the Kelly game to explore probability-and-payoff sizing, or the drawdown calculator to explore recovery arithmetic.

How this relates to Positive Skew.

Positive Skew’s sizing combines the forecast, market weight, volatility estimate and portfolio risk target. Stronger and weaker forecasts can produce different position sizes even at the same volatility estimate. Portfolio and execution constraints also matter.

The public calculator intentionally isolates one long exposure plus cash. It is not a reconstruction of the product’s forecasts, portfolio allocations or orders. The product’s built-in paper mode lets you explore a simulated workflow; simulations do not prove live performance.

Before interpreting a backtest.

• Specify the estimate, lookback, annualization convention and data available at each decision time.
• Include a lag: today’s close cannot inform an order filled earlier today.
• State exposure caps, rebalance rules and how cash or borrowing is treated.
• Include realistic costs and inspect turnover, peak leverage and drawdowns alongside average results.
• Compare equivalent risk and funding assumptions, and test periods beyond those used to choose parameters.

No backtest is presented on this page. The worked examples explain a rule; they are not evidence of excess returns or a safer investment outcome.