Sharpening Shapley Allocation: from Basel 2.5 to FRTB

Battle Count: 60

The paper is primarily focused on regulatory capital allocation and risk management within financial institutions rather than direct trading strategy development. However, it is highly relevant for quantitative risk managers, trading desk P&L attribution, capital-aware trading decisions (P&L/capital ratio optimization), and understanding how risk is decomposed across portfolios. The Shapley allocation provides a theoretically sound basis for evaluating trading desk performance and identifying hedging opportunities. The FRTB application is directly relevant to banks' trading book management.

The Shapley Monte Carlo algorithm (Algorithm 1) is clearly specified and implementable. However, production deployment requires: (1) efficient risk measure computation infrastructure (VaR, sVaR, ES, SBA), (2) handling of large-scale portfolio hierarchies (hundreds of thousands of instruments), (3) integration with existing risk systems, (4) managing computational resources for Monte Carlo simulations, (5) implementing multi-level allocation logic, (6) handling negative allocations pragmatically, and (7) ensuring data infrastructure supports scenario-level P&L access for marginal contribution calculations. The mathematical framework is well-defined but the engineering challenge in a large bank is substantial.

The paper provides detailed mathematical formulations, Algorithm 1 for Shapley Monte Carlo, and comprehensive toy-case examples with known analytical benchmarks. However, the realistic trading book data (hundreds of thousands of instruments across three hierarchical levels) is proprietary to Intesa Sanpaolo and not publicly available. The Basel 2.5 and FRTB regulatory formulas are publicly documented. The computational environment (Matlab 2022, Intel Core i5-8365U) is specified. Reproducing the toy cases is feasible; reproducing the realistic applications requires proprietary data.