historical-analog-finderlisted

# historical-analog-finder You hand over K (default 20) and horizon_days (default [30, 60, 90, 252]) and get back the K historical periods most similar to today's market regime, plus the forward SPY return distribution across those analogs. Regime-conditional forecasting. The mean is not a point forecast; the IQR is the honest read. ## When to invoke - The operator asks "what usually happens after a setup like this", "any historical analogs to now", "regime analog" - Portfolio-review workflow: after market-regime tells you WHAT, this tells you WHAT USUALLY FOLLOWS - Sizing / cash-management decisions where forward return distribution matters more than a point estimate ## What you need - `MASSIVE_API_KEY` (Stocks Starter). One SPY range-aggs call for the full history window (default 20 years). ## What you get back **Layer 1 JSON** matching [`output-schema.json`](./output-schema.json). Current regime feature snapshot (raw + z-scores), the K analogs with distance + forward returns per horizon, forward return distribution stats per horizon (p10, p25, median, p75, p90, mean, hit rate > 0). **Layer 2 rendered note**. Current-regime snapshot block, forward- distribution table across horizons, top-analog date list with per- horizon returns. See [`references/rendering.md`](./references/rendering.md). ## How it works 1. **Fetch SPY** over `history_years` (default 20). 2. **Compute a daily feature panel**: for every day with enough history (>= 260 bars), build a 9-