Relevance
9/10
Directly addresses high-dimensional portfolio construction, a core problem in quantitative trading. The PS2/FPS2 framework provides a practical two-step approach for selecting assets from large universes (tested on S&P 500 with up to 968 stocks). The factor-augmented version (FPS2) is particularly relevant for practitioners who use factor models. The paper provides concrete implementation guidance including cross-validation for Lasso parameter selection, rolling-window portfolio construction, and transaction cost modeling. Empirical results on real S&P 500 data demonstrate practical applicability, though performance is environment-dependent.
Implementation Complexity
6/10
The core methodology is conceptually straightforward (screen then estimate), but practical implementation requires: (1) Lasso regression with proper cross-validation for regularization parameter selection, (2) factor model estimation and defactoring, (3) post-screening OLS with inflated response, (4) rolling-window portfolio construction with separate screening and estimation windows, (5) transaction cost modeling. The theoretical conditions (sparsity, bounded Sharpe ratio, weak factors) require careful verification. The paper provides detailed algorithmic steps but no code. The empirical implementation involves multiple tuning parameters (J, ℓ, λ, α, τ).
Reproducibility
3/5
The paper provides detailed algorithmic descriptions (Procedures 1 and 2), explicit formulas for estimators, and specifies hyperparameter choices for simulations. However, no code repository is provided. The empirical application uses Bloomberg data (proprietary) and Ken French Data Library (public). Theoretical proofs are included in supplementary material. The cross-validation procedure for Lasso parameter selection is fully specified.