Stradion publishes directional probabilities, expected returns, and 90% intervals for a public set of large-caps. Each forecast is scored after its horizon elapses, against only the data available at the time. Directional hit rate and interval coverage are published per model and horizon, including the periods where the models do poorly.
The public signal matrix. For each symbol and horizon: a probability of a positive return, an expected return, a 90% interval, and the out-of-sample accuracy of each model to date.
A daily brief for the tracked universe, summarized from public headlines. Research only.
The research terminal behind the public surfaces: screener, model scorecards, and a signal dashboard. Access is granted by invitation.
Scored after its horizon elapses, against only the data available at the time.
Each forecast is scored after its horizon elapses, against only the data the model could have used at the time — a walk-forward, out-of-sample replay. Directional hit rate and 90%-interval coverage are published per model and horizon, including the windows where the models perform poorly. Our own analysis finds that short-horizon directional confidence does not yet rank outcomes well enough to trade on; it begins to separate around the 20-day horizon. That result is stated on the performance page. A model that is right 55% of the time is wrong the other 45%, and the published record reflects it.
No hindsight.
Per model and horizon.
Full record retained.
Stradion began as a private tool. I built the terminal I wanted for my own market work and held it to a standard I would have paid for: forecasts I could audit, uncertainty kept in the outputs, and a record I could not quietly edit. What is public here is that same tool. The methodology and the limitations are documented because that is how I use it.
The public matrix and brief stay free while the record builds.