Market Pulse turns machine-learning signals, regime context, and risk-based sizing into one clear daily read — for stocks and crypto. It never places a trade, and it never promises a profit.
No black box bravado. Each signal is the product of three transparent stages — and we tell you the model's edge is slim.
A gradient-boosted classifier over eight momentum, trend and volume features returns a calibrated probability that price rises over the next three days — not a hunch, a number.
XGBoost · isotonic calibration · p_upThe model ranks names against each other; it doesn't time the market. A breadth-and-trend gate scales conviction down when the broad tape turns hostile.
breadth · benchmark trend · scaled exposureHalf-Kelly, the 1% rule, an ATR stop, and a portfolio-heat cap turn a signal into a position you can survive. Negative expectancy sizes to zero, by design.
half-Kelly · 1% risk · 8% heat capMost trading "edges" are noise wearing a nice chart. Every theory we test runs through a leak-free, cost-aware backtest with deflated-Sharpe scoring. Most come back REJECTED — and we keep the receipts in a research ledger inside the app.
A refuted idea isn't a failure. It's the only reason to trust the ones that survive.
Liquid US equities via end-of-day data, major crypto via exchange pairs — a deliberate set kept tight so every read is fresh, not a thousand stale tickers.
Market Pulse is in private testing. It's a research and decision-support tool for people who want a sharper, more honest read — not a tip service.