Relevance
8/10
Highly relevant to quantitative trading and portfolio management. The paper directly addresses portfolio construction under heavy-tailed, asymmetric return distributions typical of technology/semiconductor ETFs. CVaR optimization, tail-risk diagnostics, and the finding that performance rankings depend on the chosen risk measure are directly actionable for quantitative portfolio managers. The comparison of mean-variance vs. tail-sensitive optimization, rolling-window rebalancing methodology, and long-only/long-short constraints are all practically relevant. However, the paper is more academic/empirical than providing a deployable trading system.
Implementation Complexity
7/10
Moderate to high complexity. Requires: (1) Hill estimator implementation with threshold selection, (2) GJR-GARCH(1,1)-t estimation with MLE, (3) Rolling FIGARCH(1,d,1) estimation, (4) CVaR optimization as linear programming (Rockafellar-Uryasev), (5) Mean-variance optimization with constraints, (6) Rolling-window portfolio rebalancing over 1008-day windows, (7) Multiple performance measures (Sharpe, Rachev, STARR). The 30-asset covariance matrix estimation and daily rebalancing add computational burden. Standard econometric packages (R, Python) can handle most components, but FIGARCH estimation and CVaR optimization require specialized libraries.
Reproducibility
3/5
Data sources are specified (Bloomberg for ETF prices, U.S. Treasury for risk-free rates). Methods are well-described with equations. However, no code repository is provided, and Bloomberg data requires a subscription. Rolling window parameters (4-year/1008-day) and bandwidth selection (m=n^0.5) are specified. The 30-ETF universe is fully listed in Appendix A. Reproducibility is moderate due to proprietary data access requirements.