5. Prediction Flow in Detail

Step-by-step process#

1. User submits a token (address or ticker).
2. System pulls the latest candle history and metadata.
3. Context Reconstruction Module builds an internal representation of recent price action.
4. Regime Detection Module classifies the current environment.
5. Primary Flow Model generates the forward path (prediction candles).
6. Risk Parameterization Layer computes:Directional bias · Target price / market-cap level · Forecast high and low · Invalidation level
7. Explanation Layer produces a concise rationale.
8. Final output is returned to the user.

Younger tokens automatically receive shorter forecast horizons to reflect higher uncertainty and faster regime changes.