Volatility Scaling for Position Sizing in Systematic Trading

In systematic trading, volatility scaling—adjusting the size of each position inversely to its current or expected volatility—is the mechanism that keeps portfolio risk constant even as market conditions shift. Without it, a strategy that holds the same number of shares in calm and crisis regimes will blow up during volatility spikes. With it, a strategy smooths drawdowns and compounds more reliably.

The Case for Scaling Position Size to Volatility

Suppose you run a mean-reversion strategy on two stocks: Stock A is typically 10% annualized volatility, Stock B is typically 30%. If you hold the same dollar amount in both, a 3% daily move in Stock A is unremarkable; a 3% daily move in Stock B is a major event. Yet both represent the same number of standard deviations of daily return.

Over long holding periods, the portfolio’s daily risk comes almost entirely from Stock B. Stock A contributes almost nothing. The strategy is implicitly concentrating risk in the volatile instrument.

Volatility scaling fixes this: size down Stock B and size up Stock A so that each contributes roughly equal expected volatility (or expected loss under your signal). Now the portfolio’s risk is truly diversified.

This becomes critical during regime shifts. In 2022, many equity long strategies saw volatility double from 12% to 24%. A fixed-size portfolio suddenly faced twice the drawdown per signal. A volatility-scaled portfolio naturally cut positions in half (roughly) as realized volatility climbed, keeping the expected loss per trade constant.

Measuring Volatility for Scaling

The most common approach is realized volatility: the standard deviation of recent daily returns over a rolling window, typically 20 to 60 days.

$$\text{Volatility}{t} = \sqrt{\frac{1}{n} \sum{i=1}^{n} (r_{t-i} - \overline{r})^2}$$

Where $r$ is daily return and $n$ is the lookback window. A 20-day rolling volatility updates daily and responds to recent market moves; a 60-day window is smoother and slower to rise or fall.

Implied volatility (from option prices) is an alternative, forward-looking measure. It reflects what the market expects volatility to be, not what it was. Some strategies prefer implied vol because it is more responsive to anticipated spikes (earnings, Fed decisions) than realized vol in hindsight.

GARCH models estimate conditional volatility by weighting recent large moves more heavily. They adapt faster to volatility clusters (days of high volatility tend to follow each other) than simple rolling standard deviation.

For most retail and small institutional implementations, 20–30 day rolling realized volatility is sufficient and transparent. The exact choice matters less than applying it consistently.

The Scaling Formula

The simplest form:

$$\text{Position Size} = \frac{\text{Target Volatility}}{\text{Current Volatility}} \times \text{Base Size}$$

If your target is 15% annualized vol per position, current vol is 30%, and your base size would be 100 shares, you hold 50 shares (0.15 / 0.30 × 100).

Some strategies use leverage limits. If the formula suggests size up to 200 shares to keep vol at target during calm periods, you cap it at 150% of base size to avoid excessive leverage. Conversely, if vol spikes and the formula suggests 10 shares, you might have a minimum position of 20% base size to keep the strategy tradeable.

Practical Considerations

Lag in volatility updates: A 20-day window takes 20 days to fully incorporate a new volatility regime. If volatility spikes on day 1, it takes until day 20 for the rolling measure to “see” it fully. Strategies living with this lag might prefer shorter windows (10 days) or GARCH to respond faster, accepting noisier signals.

Scaling too aggressively: If you scale position size to zero (or near-zero) during the lowest-volatility periods, you have almost no capital deployed when vol is calmest—and calm periods often produce steady alpha. Some strategies apply a floor: “never scale below 50% of base size” to maintain a minimum trading presence.

Cross-asset scaling: In a multi-instrument portfolio, you can scale each instrument independently (relative to its own vol) or scale all instruments to a constant target. Both are valid; the latter is often called risk parity and requires more careful monitoring to avoid concentration.

Interaction with Lookback Period and Holding Period

A strategy with a 20-day signal lookback, 20-day position holding period, and 20-day volatility window is internally consistent. All three are at the same frequency. If you stretch the signal lookback to 60 days but keep volatility scaling on 20 days, you are mixing frequencies—the signal lags the volatility measurement.

Longer-term trend strategies often use longer volatility windows (60 days) to avoid over-scaling on intra-week noise. Short-term mean-reversion strategies prefer shorter windows (10 days) to respond to fleeting mispricing.

Drawdowns and Consistency

A major benefit of volatility scaling is smoother equity curves. A fixed-size strategy on a mean-reversion signal might have a 15% drawdown in calm times (many small whipsaws) and a 40% drawdown in crisis (few large losses). A scaled strategy stretches them toward the middle—maybe 20% in calm, 30% in crisis—because it sizes down before the crisis is fully realized and sizes back up gradually.

This is not market-timing (you do not know when crisis comes), but it is crisis-aware. Volatility scaling is a form of dynamic hedging embedded in position sizing.

See also