How Options Desks Use Volatility Probability Signals to Adjust Vol Surface Positioning

Options desks quoting metals vol surfaces face a persistent information problem: implied volatility reflects the market's consensus, but it reacts to regime shifts with a lag. Probability-based volatility signals, generated from machine learning models processing alternative data, offer a way to front-run those surface adjustments rather than chase them.

This post examines the mechanics of how vol desks integrate discrete volatility probability levels into surface management, from tenor-specific recalibration to skew positioning and gamma budgeting.

From Probability Levels to Surface Repricing

A volatility probability signal classifies the likelihood that realised volatility over a defined window will exceed a threshold. The Volterra model, for example, produces 7-day, 14-day, and 30-day forecasts at five risk levels: LOW, MODERATE, ELEVATED, HIGH, and EXTREME. Each level corresponds to a probability band derived from an XGBoost classifier with a mean AUC of 0.815, walk-forward cross-validated to avoid lookahead bias.

For an options desk, the mapping from probability level to surface action is direct. Volatility probability signals at the ELEVATED or HIGH level for a specific tenor indicate that the current implied volatility surface likely underprices realised vol over that horizon. The desk response is to widen bid-ask spreads on at-the-money options at the corresponding tenor, shift the ATM vol mark higher, or reduce net short gamma exposure. Options desks typically widen bid-ask spreads on metals options when volatility probability signals shift to ELEVATED or above. Conversely, a sustained LOW signal across all three horizons suggests the surface is rich relative to likely realised outcomes, creating opportunities to sell vol or compress quoted spreads.

The multi-horizon structure matters. A signal that is HIGH at 7 days but MODERATE at 30 days describes a transient shock expectation, not a regime change. This term structure of probability drives tenor-specific adjustments rather than parallel shifts. For a deeper treatment of how horizon selection shapes risk management, see how forecast window selection shapes metals risk.

Skew Adjustment and Wing Pricing

Beyond ATM levels, volatility probability signals inform skew and wing positioning. Metals vol surfaces exhibit persistent skew patterns driven by supply disruption asymmetry: cobalt and nickel, for instance, carry structurally steeper call skew due to geographic concentration risk. Cobalt and nickel options carry structurally steeper call skew due to supply geography concentration in the DRC and Indonesia.

When a volatility signal shifts to HIGH for a specific mineral, the desk must decide where on the surface the repricing pressure will concentrate. For minerals with high Herfindahl-Hirschman Index scores, the repricing typically loads into out-of-the-money calls, steepening the skew. HHI concentration data, which feeds directly into the Volterra model's feature set, helps quantify whether a given signal reflects broad market stress or supply-specific tail risk.

Metals options desks use HHI-informed volatility signals to differentiate supply-driven skew repricing from broad market stress. A desk running short 25-delta calls on nickel, for example, would treat a jump from MODERATE to HIGH on the 14-day horizon as a direct prompt to cover or roll that exposure. The signal does not predict direction, but it quantifies the probability that the magnitude of moves will exceed what the current surface prices. Volatility probability signals quantify the likelihood that realised moves will exceed what the current implied surface prices.

Gamma Budgeting and Inventory Management

Options desks manage gamma as inventory. Net short gamma generates carry through theta decay but exposes the book to realised vol exceeding implied. Net long gamma costs carry but profits from outsized moves. The daily decision of how much gamma to hold, and at which tenors, is the core inventory problem.

Volatility probability signals provide a structured input to this decision. A desk running a daily gamma budget can map probability levels to target gamma positions: reduce net short gamma when 7-day probability reaches ELEVATED, flatten when it reaches HIGH, and consider net long gamma at EXTREME. Desks running daily gamma budgets can map Volterra probability levels directly to target gamma position thresholds.

This framework replaces ad hoc judgment with a repeatable signal. The Volterra dataset, processing 96 GDELT GKG news files daily alongside supply chain and geographic concentration features, provides the informational edge that makes this mapping actionable. The systematic processing of alternative data through the Volterra pipeline captures regime shift signals before they fully transmit into implied surfaces. Full historical backfill is available on AWS Data Exchange for backtesting these mappings against your own book.

Practical Integration

For desks evaluating integration, the key consideration is latency. Volterra signals are daily, which aligns well with end-of-day surface marking and next-day quoting decisions. Intraday scalpers need faster inputs; weekly risk reviews need less granularity. The daily cadence sits in the operational sweet spot for most metals options market-making desks.

The probability-based framework also complements existing inputs. It does not replace SABR or SVI parameterisations; it informs the calibration anchors those models require. When the volatility signal diverges from the price forecast, the desk has a quantified basis for adjusting surface parameters independent of directional views.

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