High-level summaries of the quantitative methodology behind the bi-weekly signal pipeline. Free and open — no login required.
We describe a 9-factor adaptive ensemble combining Ridge, ElasticNet, and Gradient Boosting with James-Stein shrinkage and VIX regime detection. The model adapts its weighting across regime transitions. Hypothetical results only — not indicative of future performance.
Options backtests that ignore execution costs dramatically overstate profitability. We build a per-ticker, per-moneyness spread model trained on real options chain observations. Incorporating realistic spread drag eliminates a significant fraction of apparently profitable strategies.
We propose a portfolio construction framework using drawdown-adjusted Kelly sizing with conviction tiers. Higher-conviction positions receive options overlays for convexity, while strategy selection is driven by the relationship between model-implied and market-implied volatility.
We compare model-implied volatility from GBM simulation against market ATM implied volatility to identify systematic mispricings. This vol-edge metric informs hypothetical strategy selection across conviction tiers.
Sector Momentum Decomposition and Factor Timing Signals
Q2 2026Earnings Surprise Integration with Fundamental Factor Models
Q2 2026Cross-Asset Regime Classification Using Credit Spreads and Yield Curves
Q3 2026All model outputs are time-stamped before the out-of-sample evaluation period begins. No post-hoc fitting.
Full pipeline parameters, data sources, and model specifications are documented in the open repository.
Every backtest includes evidence-based execution costs from real options chain observations. No frictionless assumptions.
See the latest bi-weekly signal snapshot.
Signal DashboardAll research is for informational and educational purposes only. Model outputs and performance figures are hypothetical and do not represent actual trading. Not investment advice. Past performance is not indicative of future results. Full disclaimer.