Entropic Value-at-Risk portfolio optimization for tempered stable Lévy processes

Battle Count: 77

Highly relevant to quantitative trading and portfolio management. The paper provides a complete parametric framework for EVaR-based portfolio construction with heavy-tailed return models, directly applicable to sector rotation strategies, risk budgeting, and tail-risk-aware allocation. The empirical backtest on liquid ETFs with transaction cost analysis makes findings immediately actionable. The ICA-based approach offers an alternative to factor models for capturing dependence structure. However, the high turnover of ICA portfolios and the need for careful parameter estimation limit practical implementation without additional engineering.

High implementation complexity. Requires: (1) fitting CTS/NTS distributions via ECDF-MSE estimation, (2) implementing MNTS parameter estimation with shared subordinator, (3) implementing FastICA and component-level distribution fitting, (4) deriving and evaluating cumulant-generating functions for each process family, (5) implementing the numerical EVaR minimization (Algorithm 1 with Brent's method over partitioned domains), (6) solving convex minimum-EVaR programs, (7) implementing Dinkelbach iterations for E-STAR, (8) implementing multi-start optimization for E-Rachev, (9) computing VaR/CVaR via Gil-Pelaez inversion and FFT for benchmark comparisons, (10) constructing rolling backtest infrastructure with turnover and transaction cost tracking. The mathematical derivations are non-trivial, particularly the weight-dependent admissible MGF domains.

The paper provides detailed mathematical derivations, algorithm pseudocode (Algorithm 1), parameter estimation procedures, and comprehensive empirical tables. However, no code repository is mentioned. The methodology relies on standard estimators (ECDF-MSE, FastICA) and well-documented distributions (CTS, NTS, MNTS). The backtest uses publicly available Yahoo Finance data for State Street Select Sector SPDR ETFs. Reproduction would require implementing the cumulant functions, EVaR optimization, and portfolio construction from scratch.