7-Day vs 14-Day vs 30-Day Volatility Horizons: How Forecast Window Selection Shapes Metals Risk Management
The choice of forecast horizon is not a modelling convenience. It is a design decision that determines which risk processes a volatility signal can actually inform. A 7-day probability forecast and a 30-day probability forecast built on the same underlying features will diverge in signal behavior, autocorrelation structure, and practical utility. Understanding these differences is prerequisite to integrating any multi-horizon volatility product into a trading or risk management workflow.
Why Horizon Length Changes the Nature of the Signal
Short-horizon volatility forecasts capture fast-moving, event-driven risk. A 7-day volatility window is dominated by immediate catalysts: inventory report surprises, tariff announcements, exchange rule changes, or supply disruption headlines. The signal-to-noise ratio on geopolitical and news-driven features is highest at the 7-day horizon. In the Volterra model, GDELT-derived news intensity and tone features carry their strongest predictive weight at this shortest window, reflecting the rapid decay of information shocks in liquid metals markets.
A 30-day volatility forecast, by contrast, integrates slower-moving structural factors. Geographic supply concentration, measured via Herfindahl-Hirschman Index scores, exerts more influence at the 30-day horizon because supply disruptions take weeks to propagate through physical delivery chains. The 30-day window also captures roll-cycle effects on LME contracts and options expiration clustering on COMEX, both of which inject periodic volatility that washes out at shorter horizons.
The 14-day horizon occupies an intermediate zone. It blends event-driven sensitivity with early structural signals. 14-day volatility forecasts are most useful for hedging programs that rebalance biweekly or for options desks managing positions with two to three weeks until expiry.
Mapping Horizons to Risk Management Workflows
Each horizon aligns to a distinct operational cadence:
7-day forecasts serve intraweek delta hedging, short-dated options positioning, and real-time risk limit monitoring. When a 7-day signal moves from MODERATE to HIGH for copper on COMEX, the immediate implication is that realized vol over the coming week has an elevated probability of exceeding recent norms. Options desks use this to adjust vol surface positioning on near-term tenors. 7-day volatility forecasts are most responsive to news-driven catalysts because information shocks in metals markets typically decay within five to ten trading sessions.
14-day forecasts align with biweekly hedge rebalancing and medium-term spread positioning. For systematic strategies, the 14-day window often provides the best tradeoff between forecast stability and lead time. Signal transitions at the 14-day horizon tend to be smoother than at 7 days, reducing whipsaw risk in rule-based systems.
30-day forecasts inform monthly VaR recalibration, procurement contract timing, and portfolio-level risk budgeting. A 30-day ELEVATED or HIGH reading on nickel, for example, signals that the coming month's realized vol distribution is skewed upward, which has direct implications for supplier contract structuring and inventory buffer decisions. The 30-day horizon captures structural supply concentration risk that shorter windows miss.
How Multi-Horizon Signals Interact
The most actionable information often comes from horizon divergence. When the 7-day signal for a mineral is HIGH but the 30-day signal remains MODERATE, the model is detecting a transient event-driven shock without structural follow-through. Conversely, when the 30-day signal escalates before the 7-day signal, it suggests a slow-building risk that has not yet been priced into near-term implied vol. Multi-horizon divergence between 7-day and 30-day forecasts often precedes sustained volatility regime shifts in metals markets.
The Volterra pipeline produces all three horizons daily for each of the 12 covered minerals across LME, COMEX, NYMEX, and SGX. The model's walk-forward cross-validation framework ensures that each horizon is validated independently, avoiding lookahead contamination between windows. With a mean AUC of 0.815 across horizons, the classification performance is stable, though predictive feature importance shifts materially between the 7-day and 30-day outputs.
Practical Considerations for Horizon Selection
There is no universally optimal horizon. The right choice depends on portfolio turnover, hedging frequency, and the operational latency of your risk process. A desk that can act within hours of signal delivery will extract the most value from 7-day forecasts. A procurement team operating on monthly cycles should anchor to the 30-day output and use the 7-day signal as a tactical overlay.
For systematic strategies, combining all three horizons as features in a downstream allocation model can capture both the speed of event-driven signals and the persistence of structural risk. The Volterra dataset provides all three horizons in a single daily delivery, structured for direct ingestion. Figures from the Volterra daily pipeline. Full historical backfill available on AWS Data Exchange.
Horizon selection is a design decision with direct PnL consequences. Choosing the wrong window means either reacting too late to fast shocks or overtrading on noise that the market absorbs within days. The multi-horizon approach exists precisely because no single window serves all risk functions simultaneously.