Continuous Timing Signals for Growth–Defensive Style Allocation: Factor Attribution, Risk Matching, Out-of-Sample Evidence, and a Bond/Credit Incremental Extension

Battle Count: 149

Highly relevant to quantitative trading as it provides a transparent, implementable framework for dynamic style allocation between growth and defensive assets. The continuous score approach avoids arbitrary regime labels, the walk-forward validation provides realistic out-of-sample evidence, and the bond/credit incremental extension demonstrates practical overlay methodology. The explicit transaction cost modeling and turnover analysis make it directly applicable to live trading. However, the inability to beat 100%G in raw CAGR and the high turnover (~470%) limit its practical edge in a strong growth market.

Moderate complexity. The methodology uses well-defined mathematical operations (z-score normalization, softplus, tanh, EWMA) that are straightforward to implement. The main complexity lies in the parameter grid search (793 configurations for the bond/credit branch), the walk-forward validation framework, and the careful handling of signal timing (signals formed after close, applied next day). No deep learning or complex optimization is required. The GitHub repository provides full implementation code.

Code, data artifacts, and reproducible empirical materials are explicitly available on GitHub. The paper provides detailed parameter grids, exact formulas for all components (softplus, tanh mapping, EWMA smoothing), transaction cost specifications, and full benchmark comparisons. The methodology is fully transparent with no black-box components.