pocket-quant — Free-Data Quant Engine

A free-data quantitative trading research engine — macro crash-risk scoring, SEC 13F institutional-flow signals, congressional-trade aggregation, and a rule-based strategy layer producing per-account BUY/SELL/HOLD, with a Streamlit dashboard. Built entirely on free data, no paid feeds.

Blends several signals into per-account recommendations using only free data (yfinance, FRED, SEC EDGAR, QuiverQuant): a crash-risk model — a composite 0–100 score from ~16 leading macro indicators (Buffett indicator, margin debt, yield curve, Sahm rule, HY credit spreads) calibrated over 1986–2025.

An institutional 13F tracker weights ~12 filers by their *measured* forward edge (most 'famous-investor edges' don't survive proper sample sizes — the weights reflect that), plus congressional buy/sell pressure aggregated from STOCK Act disclosures.

A rule-based layer scores multi-factor (smart-money / quality / momentum / value / earnings) with trailing stops and drawdown control, surfaced through a Streamlit dashboard. *Personal research project — not investment advice.*