Surface-Driven Stochastic Volatility for Commodity Options: Identification of Stochastic Vol-of-Vol and Leverage from Smile Dynamics

Battle Count: 0

Highly relevant for desks trading commodity options. It provides a method to extract dynamic risk parameters (vol-of-vol, leverage) directly from observable market data (CVOL) rather than relying on static calibration. The neural network surrogate enables real-time pricing and hedging, which is critical for high-frequency or large-scale portfolio management.

Moderate to High. Requires implementing stochastic differential equations, solving 2D PDEs (or running large-scale Monte Carlo), and training a neural network. The statistical estimation of the surface factors is straightforward, but the numerical pricing engine requires careful handling of boundary conditions and stability.

The paper provides detailed algorithmic descriptions for Monte Carlo, Finite Difference, and Neural Network training. It specifies data sources (CME CVOL) and parameter estimates. However, the specific code repository is not explicitly linked in the text, though the methodology is fully described.