How Options Desks Use Volatility Probability Signals to Adjust Vol Surface Positioning
From Implied Vol to Probability-Weighted Positioning
Options desks traditionally calibrate vol surfaces using a combination of realized volatility, implied volatility from active strikes, and dealer flow. In metals markets, where liquidity concentrates in a narrow band of strikes and tenors, surface construction already requires interpolation and judgment. The challenge intensifies for critical minerals where option markets are thinner and the underlying spot dynamics are driven by supply concentration, geopolitical disruption, and procurement cycles rather than purely financial flows.
Volatility probability signals add a forward-looking layer to this process. Rather than extrapolating from trailing realized vol or relying solely on the implied surface, a probability forecast assigns discrete likelihoods to volatility regimes over defined horizons. Volterra produces exactly this: 7-day, 14-day, and 30-day probability forecasts across five risk levels (LOW through EXTREME), generated daily from an XGBoost model with a mean AUC of 0.815. Volterra's XGBoost model processes 96 daily GDELT news files alongside supply chain and geographic concentration features to generate volatility probability forecasts. The output is not a point estimate of future vol; it is a probability distribution over volatility states, which maps more naturally onto how desks think about surface adjustments.
Mapping Probability Levels to Surface Adjustments
When a signal shifts from MODERATE to ELEVATED for a given mineral and horizon, the desk faces a concrete question: where on the surface does this information land, and how much weight does it deserve relative to existing implied levels?
Options desks using volatility probability signals typically adjust the at-the-money vol level, the skew slope, and the term structure independently. An ELEVATED 7-day signal with a MODERATE 30-day signal, for instance, suggests a short-dated vol premium without a corresponding shift in longer tenors. This steepens the front end of the term structure. Conversely, an ELEVATED 30-day signal paired with a LOW 7-day reading may indicate that the market has not yet priced an emerging structural risk, creating opportunity in calendar spreads. Selecting the right forecast horizon shapes how desks allocate risk across the term structure.
Probability signals are especially useful for skew adjustments in metals with asymmetric supply risk. Cobalt and lithium, where geographic concentration is extreme, tend to exhibit sharper upside skew during supply disruptions. A rising probability of HIGH or EXTREME volatility in these minerals gives the desk a quantitative basis for bidding up out-of-the-money call implied vols, rather than relying on discretionary assessment of headline risk. Supply concentration, measured by the Herfindahl-Hirschman Index, directly quantifies the tail risk embedded in these minerals.
Integrating Signals into Gamma and Vega Management
Beyond surface repricing, probability signals inform how desks manage their Greeks dynamically. Metals options desks adjust gamma exposure based on forward-looking volatility probability rather than trailing realized vol alone. A transition from LOW to ELEVATED on a 7-day horizon implies the desk should reduce net short gamma or add long gamma hedges, even before realized vol confirms the move. Waiting for realized vol to catch up means accepting the P&L drag of being short gamma through the acceleration phase.
Vega management follows a similar logic but operates across tenors. Volterra's multi-horizon output lets desks decompose vega exposure into buckets that align with the signal's probability profile. If the 14-day and 30-day signals diverge, the desk can construct vega-neutral positions within a single mineral while maintaining directional vol exposure at the horizon where the signal is strongest. Volterra's multi-horizon framework allows desks to construct vega-neutral positions at one tenor while maintaining directional vol exposure at another.
Practical Considerations for Signal Integration
Probability signals from any model, including Volterra, should be treated as one input in a multi-factor surface management process. The Volterra model's walk-forward cross-validation framework, detailed in our methodology discussion, ensures out-of-sample integrity, but no single signal replaces the desk's view on microstructure, dealer positioning, or event calendars.
The most effective integration pattern treats the probability output as a prior that the desk updates with real-time flow and positioning data. When the model's prior aligns with observed flow, the desk can size adjustments with higher conviction. When they diverge, the divergence itself becomes informative, often flagging that the market is discounting a risk the model has identified, or vice versa.
Critical minerals vol surfaces are structurally harder to maintain than base metals because option liquidity is sparse and reference prices are less transparent. Probability-based signals partially compensate for this by providing a systematic, repeatable assessment of forward volatility states that does not depend on option market depth.
Figures from the Volterra daily pipeline. Full historical backfill available on AWS Data Exchange. The complete Volterra dataset covers 12 exchange-traded critical minerals across LME, COMEX, NYMEX, and SGX, delivering the probability inputs that desks need for daily surface calibration.