Volatility targeting: how to calculate position size from volatility

The one-line version

Decide how much the position should move on a typical day — your target volatility — and scale your size by the ratio of that target to what the instrument is currently doing: position scale = target volatility / current volatility.

If you want a position that moves about 1% on a typical day and the instrument's recent daily move is 1%, you trade at your base size. If recent daily moves are 2%, you trade at half base size. If they are 0.5%, you trade at double. The point is not the specific numbers; it is that the risk you take stops depending on which week you happen to trade.

The same logic applies whether your sizing unit is lots, contracts or account percentage. Volatility targeting is a multiplier applied on top of whatever base size your risk rule already produced.

Why a fixed stop is not fixed risk

Suppose your rule is a stop 50 points away, on an instrument that typically moves 50 points in a day. The stop sits about one day's move from entry. Now suppose the instrument's typical daily move rises to 150 points and you keep the same 50-point stop and the same size. The stop is now a third of a day's move. You will be stopped out by ordinary noise, and the sequence of small losses will do more damage than the original risk implied.

The usual fix at that point is to widen the stop, which increases the loss per trade, so you also reduce the size to keep the money at risk constant. That is volatility targeting arrived at from the stop side. It is the same calculation as scaling by the volatility ratio, just expressed in points instead of in percent.

Formally, dollars at risk per trade is approximately size × volatility. Hold that product constant and you have constant risk, which is what you actually wanted when you wrote down your position size rule.

Choosing the volatility estimate is the whole problem

The formula is trivial; the estimate is not. You need a measure of "current" volatility, and every choice of lookback window trades one failure for another.

A short window, say 10 bars, reacts fast. It cuts your size within days of a volatility spike, which is what you want, but it is noisy: a couple of large bars can halve your size and then hand it straight back, and the whipsaw costs you transaction costs and missed exposure.

A long window, say 100 bars, is stable but slow. It will keep you at full size through the first days of a genuine regime change, which is exactly when the damage happens.

A common compromise is a window somewhere in the 20 to 60 bar range, or an exponentially weighted estimate that decays older observations. The honest position is that this is a trade-off with no dominant answer, so it should be chosen to match how quickly your strategy's losses accumulate, and then left alone.

Two practical guardrails are worth more than fine-tuning the window. Cap the multiplier in both directions — a floor so that a quiet market does not lever you up without limit, and a ceiling so that a spike does not drive your size to zero. And never let the multiplier override the position-size rule that came from your risk tolerance; it can only reduce or modestly increase what that rule already allowed.

Three ways it can still hurt you

First, it does not create return. Volatility targeting changes the distribution of outcomes and the size of drawdowns. A strategy with no edge is still a strategy with no edge, and it will now lose money at a steadier rate. If you find a study showing volatility targeting improves a risk-adjusted measure, the improvement generally comes from drawdown reduction rather than from higher average return.

Second, it can cut exposure at exactly the wrong moment. Trend-following strategies often make their money from large moves, and large moves come with high volatility. Scaling down because volatility rose means scaling down just as the profitable part of the move begins. This is a real and well-documented tension with momentum strategies, and it is a reason to test the rule on your own returns rather than adopting it because it sounds prudent.

Third, it depends on tradeable liquidity. In a fast market, spread widens and slippage grows, so the realised risk of a position is higher than the estimate your multiplier was based on. On thin instruments the volatility estimate itself becomes unreliable precisely in the conditions where you are relying on it.

How to test whether it helps your strategy

Run the same strategy, same window, same costs, twice: once with fixed size and once with the volatility multiplier. Compare the trade list, not just the summary — entries, exits and sizes will differ, and the differences are where the answer is.

Then compare three things specifically. Maximum drawdown, because that is what the method is supposed to improve. Return, because it is what you are paying for the improvement. And the correlation of its returns with the ones from the fixed-size version, because if they are near-identical you have added a parameter without adding an effect.

Test more than one window length and pick by out-of-sample behaviour, not by the best in-sample number. With a short sample, the best window length is usually the one that happened to fit — this is the same multiple-testing trap that applies to any tuned parameter, and it is easy to walk into because the change feels like risk management rather than optimisation.

What this does not do

It does not make a fixed risk per trade true in the tail. It equalises typical risk. Gaps, halts and weekend jumps bypass the stop entirely, and no scaling rule can size for an event that has not happened yet.

It does not know the difference between volatility that is dangerous and volatility that is profitable. It only knows a number went up.

And it is not a substitute for a risk limit at the portfolio level. If every position is individually volatility-targeted but they are all correlated with each other, your total risk still scales with the number of positions, which is a separate problem with a separate fix.