Large-Scale Asset Selection via Metric Dependence with Enriched High Frequency Information

Battle Count: 82

Highly relevant to quantitative trading. The paper directly addresses large-scale asset selection for portfolio construction, a core problem in quantitative finance. It provides a practical two-stage pipeline (screening + optimization) applicable to real trading universes. The use of intraday volatility curves as risk state information is directly actionable for risk-adjusted allocation. The empirical study on 2938 Chinese A-shares demonstrates practical feasibility. The Sharpe ratio improvements (1.22 vs 0.70 for D-SEVIS at d=30) are economically meaningful. However, the lack of transaction cost modeling and the specific market (Chinese A-shares) somewhat limit direct applicability to other markets.

Moderate to high complexity. Requires: (1) high-frequency data processing to estimate intraday spot volatility curves on a grid, (2) computation of pairwise distances in the point-curve metric space, (3) empirical Fréchet mean estimation (iterative or closed-form for L2), (4) target space partitioning and within-cell Fréchet variance computation, (5) rolling window implementation for backtesting. The L2 metric makes Fréchet means computationally tractable (componentwise averages), but the overall pipeline involves multiple estimation steps. Theoretical tuning parameters (partition size Mn, threshold κn^{-τ}) require careful selection.

The paper provides detailed mathematical formulations, algorithmic steps, and simulation parameters. However, no GitHub repository or code is mentioned. The empirical study uses proprietary Chinese A-share high-frequency data (2938 stocks, July 2023 to December 2025) which may not be publicly accessible. Theoretical proofs are fully provided. Simulation settings are well-specified with 500 replications.