7-Day vs 14-Day vs 30-Day Volatility Horizons: How Forecast Window Selection Shapes Metals Risk Management

Volatility is not a single number. The same mineral can register LOW risk at 7 days and ELEVATED risk at 30 days, or vice versa. Choosing the wrong horizon for your decision cycle means either reacting to noise or missing the regime shift entirely. The Volterra model produces probability forecasts across all three windows precisely because no single horizon serves every use case.

Why Horizon Length Changes the Signal, Not Just the Timeframe

A 7-day volatility forecast and a 30-day volatility forecast are not the same measurement stretched across different calendars. They capture structurally different dynamics. Short-horizon volatility in metals is dominated by order flow, inventory reports, and breaking news. Longer-horizon volatility reflects supply chain disruptions, policy shifts, and structural demand changes.

Seven-day realized volatility in metals markets typically exhibits higher kurtosis than 30-day realized volatility. This is because short windows capture individual shocks before mean reversion dampens them. A single LME warehouse stock drawdown can spike 7-day nickel volatility while leaving the 30-day measure largely unchanged. Conversely, a gradual tightening in cobalt supply from the DRC may not register at 7 days but steadily lifts the 30-day risk profile.

The Volterra pipeline processes 96 GDELT GKG news files daily alongside geographic concentration and supply chain features. At shorter horizons, the news flow features carry more predictive weight. At 30 days, the supply concentration metrics and structural factors dominate. The model's walk-forward cross-validation ensures these feature importance shifts are captured honestly across regimes.

Matching Horizon to Decision Cycle

Each forecast window maps to a distinct set of trading and risk management decisions.

7-day forecasts serve options desks managing weekly expiries and delta hedging books. Seven-day volatility probability forecasts are most actionable for desks managing weekly options expiries and intraday delta hedging. When a 7-day signal moves from MODERATE to HIGH on copper, that is a direct input to short-dated vol surface repricing. For options desks adjusting vol surface positioning, the 7-day window aligns with the gamma exposure they need to manage now.

14-day forecasts bridge the gap between tactical and strategic. Fourteen-day volatility forecasts serve systematic traders running momentum or mean-reversion strategies with biweekly rebalancing cycles. This horizon captures event clusters: a sanctions announcement followed by retaliatory export controls, or sequential inventory reports that confirm a trend rather than a one-off print.

30-day forecasts align with monthly VaR calculations, procurement hedging cycles, and portfolio risk reviews. Thirty-day volatility horizons align with monthly VaR calculations and procurement hedging windows used by industrial consumers. A procurement team sourcing lithium carbonate needs to know whether the next month's price distribution is widening, not whether tomorrow will be choppy. This is also the horizon where structural supply chain risk becomes measurable rather than speculative.

Horizon Divergence as a Signal Itself

When short and long horizons disagree, the divergence itself carries information. A mineral showing HIGH at 7 days but LOW at 30 days is experiencing a transient shock that the market expects to resolve. A mineral showing LOW at 7 days but ELEVATED at 30 days suggests a slow-building structural risk that has not yet hit spot pricing.

Divergence between 7-day and 30-day volatility forecasts for the same mineral functions as a term structure signal for vol desks. Options traders recognize this pattern as analogous to the vol term structure: backwardation in vol (short higher than long) implies mean reversion; contango in vol (long higher than short) implies unresolved risk ahead. The Volterra dataset captures these cross-horizon relationships daily across 12 exchange-traded critical minerals.

Practical Implementation

Running all three horizons simultaneously is not redundant. Multi-horizon volatility monitoring across 7-day, 14-day, and 30-day windows reduces the risk of horizon mismatch in hedging and position sizing. A risk manager using only 30-day forecasts will underreact to short-lived but violent moves. A trader using only 7-day forecasts will miss the slow buildup of a supply disruption that eventually reprices the entire forward curve.

The Volterra model's five risk levels (LOW, MODERATE, ELEVATED, HIGH, EXTREME) apply independently at each horizon. This means a single mineral can carry three distinct risk classifications simultaneously. The XGBoost architecture, with a mean AUC of 0.815 across walk-forward validation, maintains calibration at each horizon independently rather than interpolating between them.

For desks integrating these signals, the recommendation is straightforward: use 7-day for execution and hedging, 14-day for position management, and 30-day for portfolio-level risk allocation. The interaction between horizons tells you whether you are managing a spike or a regime change.

Figures from the Volterra daily pipeline. Full historical backfill available on AWS Data Exchange.