Volatility targeting in a multi-strategy book

Volatility targeting is the single most reliable risk-management practice in systematic investing. The literature has covered the single-strategy case thoroughly. The multi-strategy case is more interesting, because the practical question is no longer 'what is the volatility of this strategy' but 'how does correlation between strategies move with the regime that determines their joint volatility'.

Three implementation choices that matter

Per-strategy vs. portfolio targeting

Targeting volatility per strategy is the safer default — it prevents any single line of research from dominating the book. Targeting only at the portfolio level relies on stable cross-strategy correlations, which is a stronger assumption than most managers realise.

Look-back window

Shorter windows respond faster to changing volatility and reduce drawdowns when regimes shift, at the cost of more turnover. Longer windows are cheaper to run and capture longer-cycle structure. We use multiple horizons and combine them; the combination is more stable than any single choice.

Floor and ceiling

Pure inverse-volatility sizing can produce extreme leverage during quiet regimes, which is exactly when leverage is most dangerous because the quiet itself may be the precursor to a vol spike. A ceiling on gross exposure prevents the most familiar accidents.

What this looks like in practice

Volatility-sized strategies, summed at constrained portfolio weights, with explicit ceilings on gross and per-strategy exposure, behave more conservatively than the same strategies under any single-axis sizing rule. The trade-off is that they leave money on the table in trending regimes. That is a trade-off we accept.