Stradion publishes probabilistic forecasts for a public set of large-caps — direction, calibrated odds, and 90% ranges — then scores every one out of sample as its horizon matures. Probabilities, not price targets. Scored, not promised. The record, misses included, is on the page.
The public signal matrix. Per symbol and horizon: probability of an up move, a point forecast, a 90% range — and live, out-of-sample accuracy for every model.
A compact daily brief for the tracked universe, condensed from public headlines with a research-only frame.
The full research terminal behind the public surfaces — screener, model scorecards, and a deep signal dashboard. Private, granted to research collaborators.
Every forecast is scored out of sample as its horizon matures, and the result is shown whether it flatters us or not.
Each forecast is scored only after its horizon elapses, and only against data the model could have known at the time — a walk-forward, out-of-sample replay with no hindsight and no cherry-picking. We publish directional hit-rate and 90%-interval coverage per model and horizon, and we show the uncomfortable parts: where short-horizon confidence does not yet rank outcomes well enough to trade on, we say so on the page rather than bury it. A model that is right 55% of the time is still wrong 45% of the time — we would rather you see that than be sold certainty.
No hindsight.
The whole record.
Full history, exportable.
Stradion began as a private tool. Instead of designing for a hypothetical customer or chasing monetization on day one, I built the terminal I wanted for my own market work — and held it to the standard I would have paid for: forecasts I could audit, uncertainty left in, and a record I could not quietly edit. What is public here is that same tool, opened up. The honesty is not a posture; it is what the thing was built for.
The public matrix and brief stay free while the record builds.