Transparency · How alerts work
How alerts work
Read score (v2) is a reliability lens: clearer historical patterns tend to match Confirm more cleanly. It does not predict prices. Triggers translate the read into three levels you can watch.
Market price alerts can email when the stored snapshot crosses your threshold. Model(score) alerts update in the app only — we don't email them, so they're harder to mistake for buy or sell instructions.
How does the score get to a trigger?
Higher scores tend to mean fewer fake-outs after Confirm. Volatility and choppy series are penalized; clean trend structure is rewarded. The score then maps to three named levels:
- Confirm— "strength confirmed" level.
- Pullback— "preferred entry zone" range (when present).
- Invalidate— "read invalidated" level.
When does a trigger refresh?
Triggers update when the score or levels move enough to change the read. When levels move, you'll see an "Updated" banner and a short reason. Thin-history cards may omit triggers entirely rather than show a low-confidence signal.
What evidence supports this?
In a 180-day walk-forward harness, cards in the top decile by read score v2 had a higher "confirm without stop break" rate than the baseline:
Snapshot · 2025-09-12 · 1,798 variants · 33,850 evaluation rows · walk-forward harness on stored series. Not financial advice. Future conditions can differ.