BENJAMIN ZSIGRI / TRIBLOC DEMOSYNTHETIC PREDICTIONS & OUTCOMES

EVALUATION / THRESHOLDS / LABEL ELIGIBILITY

A score needs
the right denominator.

Explore false positives, missed positives and incomplete outcomes. Change the threshold or cutoff date, then inspect exactly which records were scored.

CONFIGURE A FICTIONAL EVALUATION

A positive prediction means score ≥ threshold. Only outcomes within their record's window and recorded by the cutoff enter performance metrics.

LIVE BROWSER CALCULATION

Evaluating…

CONFUSION MATRIX / SCORED RECORDS ONLY

Actual ↓ / Predicted →PositiveNegative
Positive0True positive0False negative
Negative0False positive0True negative

WHAT DID NOT ENTER THE METRICS?

Pending: window still open
0
Unlabelled: no eligible known outcome
0
Excluded: prediction after cutoff
0

Missing outcomes are never treated as negatives. Pending and unlabelled records remain visible and lower label coverage. Future predictions are excluded from the coverage denominator.

TRACE EVERY DENOMINATOR

Inspect the evaluation records.

RecordScorePredictionEligible labelStatusCell / reason
Inspect the reproducible report JSON