Explainable projections from cyclical regression + supervised models.
Hedgtrade combines harmonic decomposition, regime/seasonality cohorts, and supervised learning to publish decision-ready risk signals with drivers, reversal zones, and confidence — continuously across 2,600+ instruments.
A repeatable ML workflow that turns market structure into compact features, then outputs projections and risk states that can be discussed, governed, and acted on with consistent boundaries.
- Cyclical regression estimates oscillatory parameters and turning points.
- Supervised models map feature states → probable path distributions.
- Cohort seasonality adds context (decennial, election, day-of-month, etc.).
- Explainability keeps outputs audit-friendly: drivers, timestamps, and change notes.
What teams get out of it
Decision-ready signals
Direction, confidence, strength, and “what changed” — built for weekly cadence.
Reversal zones
Practical levels to plan TP/SL, invalidation, and hedge triggers.
Regime + cohort context
Seasonality cohorts reduce whipsaw and clarify when a signal is “in character.”
How the ML pipeline works
Built to reduce overfit risk: we separate “structure detection” from “forecasting,” then publish outputs with explainability your team can review.
Decompose & detect
Dominant harmonics + trend/momentum features summarize the state.
Filter & align
Cycles are filtered and aligned with regime + cohorts to reduce noise.
Project & explain
Supervised models output path probabilities + reversal zones with drivers.
Reversals, trends & momentum
The model publishes reversal zones for planning, alongside composite trend/momentum context — helping teams avoid reactive, narrative-driven positioning.
See it on your universe
We’ll run the workflow end-to-end on representative assets from your universe: feature state → projection → reversal zones → decision boundaries → reporting cadence.