ML and AI

• Backtest Discipline — Overfitting and multiple-testing control, data snooping, purged/embargoed walk-forward CV, the deflated Sharpe ratio, and the non-negotiable 2016–2026 ThetaData stress set. status: draft
• LLM Agent Roles — Where LLM agents genuinely fit in the options system (news triage, retrieval, structure generation, EV enforcement, journaling) and where they do not (probability estimation, price forecasting), plus the human-in-the-loop gate. status: draft
• ML for Volatility Prediction — Feature ideas from the IV surface and market activity, model classes worth trying, realistic expectations versus GARCH/HAR baselines, and the standard pitfalls. status: draft
• Volatility Forecasting Baselines — Realized-vol estimators, GARCH-family and HAR-RV models, and why any ML vol forecast must beat these baselines out-of-sample to matter. status: draft

Back to the bundle index.