One model, trained in-house, reading every symbol on every cycle. It weighs its own signal layers, scores its conviction, and returns a calibrated forecast with the reasoning behind it. When the picture is unclear it says so, instead of guessing.
Markets reward humility. The firms that survive over decades, Renaissance, Two Sigma, AQR, don't bet on one perfect indicator. They combine many weakly correlated signals into a single disciplined view. The Eltrove Engine is built on the same principle, end to end, as one model we trained and own.
Inside, the Engine reads a symbol through several signal layers at once, each looking at a different face of the market. When the layers reinforce each other, conviction is high. When they pull apart, conviction is low and the forecast says so.
For a given symbol, each signal layer produces a directional read and a confidence. No layer's read leaks into another's.
Each layer is scaled by its rolling out-of-sample accuracy. A layer that has been sharp in this regime carries more weight than one that is drifting.
Aggregate conviction must clear a regime-aware threshold. Conviction scales the forecast range and the confidence badge. Below threshold the Engine reports low conviction instead of forcing a call.
The result lands on the Forecast tab as a directional lean, a likely range, and a confidence badge, with the plain-English reasoning shown.
The Eltrove Engine is not a black-box stock picker shouting "BUY NOW."
You'll see the Engine's directional lean, the likely range around it, the confidence it assigned, the why behind the call, and where it was wrong before, because the scorecard is public. The dashboard shows the work.
Automated execution is coming once we are licensed. When it arrives you will be able to connect a broker and let the Engine act on its own calls. It is not live today, and we make no promise of returns.
Run any ticker through the Forecast tab, or create an account to keep your history.