Distribution-Based Ranking and Allocation with Performance Measures
We develop a distribution-based framework for ranking financial assets when performance measures are estimated with uncertainty and are statistically dependent across assets. The method transforms pairwise comparisons of estimated performance measures into domination probabilities, which define a stochastic matrix whose Perron-Frobenius eigenvector provides a complete ranking and, after normalization, a long-only eigenportfolio. For Sharpe ratio estimators, we initially derive closed-form domination probabilities from their joint asymptotic distribution under temporally independent and identically distributed multivariate returns. The associated covariance matrix incorporates correlations, coskewness, and cokurtosis. We further extend the framework to serially dependent returns, to alternative performance measures, including the modified Sharpe ratio and the VaR ratio, and to conditional and finite-sample Edgeworth-corrected domination probabilities. We apply the methodology to S&P 500 constituents at different frequencies and use a block-bootstrap analysis to confirm the robustness of the asset-selection and portfolio-performance results.
Herrero, C., A. León, T-M. Ñíguez and A. Villar (2026). «Distribution-Based Ranking and Allocation with Performance Measures». Social Science Research Network (SSRN) (July). http://dx.doi.org/10.2139/ssrn.6963959

