Application of Quasi Monte Carlo and Global Sensitivity Analysis to Option Pricing and Greeks

Battle Count: 61

Highly relevant for quantitative finance practitioners dealing with derivative pricing, hedging, and risk management. The findings directly impact computational efficiency of Monte Carlo-based pricing engines used in trading desks and risk management departments. The comparison of FD vs AAD with QMC is particularly relevant for implementation decisions in production systems. However, it is more focused on computational methodology than on trading strategy development per se. The speed-up factors of up to 10^3 in scenario count are significant for real-time risk computation and hedging.

QMC implementation is relatively straightforward (replacing PRN with LDS generators like SobolSeq8192). Brownian Bridge and PCA discretizations add moderate complexity. GSA computation requires D+2 function evaluations per trial, which is manageable. AAD implementation is significantly more complex, requiring either manual coding or automatic differentiation tools, and handling of discontinuous payoffs. The paper's key finding that FD+QMC can match AAD+MC in accuracy for small numbers of Greeks reduces implementation burden. Overall, the methodology is implementable but requires careful attention to sampling order, generator quality, and parameter tuning.

The paper provides detailed parameter settings (maturity, strike, barriers, volatilities, correlations, number of time steps, number of paths), specifies the SobolSeq8192 generator from BRODA, uses Matlab for computations, and provides comprehensive tables of results. However, no code repository is explicitly linked. The methodology is well-documented with formulas and algorithmic descriptions. The paper was published in Wilmott Magazine with a DOI.