Probability-model scenario runner with baseline comparison and parameter provenance.
1. Start with production
The baseline branch uses the same defaults as the live Probability page.
2. Tune one assumption
Change a density gate, regime window, or blend threshold to isolate its effect.
3. Run the scenario
The backend runs baseline and tuned configs side by side without saving changes.
4. Read the deltas
Compare quality state, conditional shift, touch-risk changes, and provenance JSON.
No scenario run yet.