Each dot is one report. A well-calibrated model shows heavy reports clustering at high scores, light at low.
Reports by species
Observation coverage by hatch. Species with few reports have less reliable calibration.
Reports by river
River coverage balance. Gaps indicate where the model is least validated.
Reports over time
Weekly submission volume. Growing means the community data loop is working.
Pending approvals
Reports awaiting review before publishing publicly.
Loading queue…
Loading…
User management
All registered accounts. Ban a user to block future submissions.
Email
Display name
Joined
Reports
Approved
Status
Action
Loading users…
All hatch reports
Loading…
Date / Time
River
Species
Intensity
Model score
Delta
Water °F
Reporter
Notes
Loading reports…
Precision & recall
How often does a high model score correspond to a real hatch report?
Calibration curve
Predicted score (x) vs. actual hatch rate (y). The diagonal is perfect calibration. Above = underconfident, below = overconfident.
Score distribution at report time
What scores were running when anglers actually saw hatches? Ideal: most reports cluster at 60–90.
Brier scores by species
Lower = better. 0 is perfect, 0.25 is equivalent to always predicting 50%. Needs 20+ reports per species to be meaningful.
Loading…
False positive analysis
Days where the model scored ≥65 but no report was filed. Could be correct predictions with no angler present, or genuine model over-confidence.
Signal
Reports with this signal
Avg score when present
Accuracy contribution
Analyzing signals…
Per-species calibration
How well the model predicts each species independently. Cards show avg score on report days vs. expected, and data coverage.
Loading…
Recommended weight adjustments
Based on current report data. Apply these to close the gap between predicted and observed scores.
Species
Signal
Current weight
Suggested weight
Reason
Calculating recommendations…
Signal weight editor
Adjust how much each environmental signal contributes to a species' probability score. Changes preview instantly — use "Apply & save" to persist to localStorage.
Data export
Download calibration data for offline analysis or backup.
All hatch reports (CSV)
Every submitted report with river, species, intensity, time, water temp, notes, and computed model score at time of submission.
Calibration summary (JSON)
Per-species Brier scores, precision/recall metrics, and recommended weight adjustments based on current data.
Tuned weights (JSON)
Current signal weights as configured in the weight tuner. Import this into the main site to apply your calibrations.
Danger zone
Irreversible actions — be careful.
Delete all reports
Permanently removes all angler reports from Supabase. Cannot be undone. Export first.
Import from fishing reports
Paste a URL from a fishing report site. Claude reads the page and extracts hatch observations to calibrate the model.
Known sources
Click to load URL
— or paste report text directly (use this if URL fetch fails) —