[ PROOFECONOMY ]™GlossaryOriginSpecsEfficacy Rate
Efficacy Rate is Stopped divided by Stopped plus Missed over the bad canaries in a valid Dead Canary Protocol run. It is never published without the False Block Rate, because a control that blocks everything is not efficacious.
The run metrics
| Metric | Definition |
|---|---|
| Efficacy Rate | Stopped / (Stopped + Missed) |
| False Block Rate | False block / (Passed + False block) |
| Novelty Efficacy Rate | Efficacy Rate over mutated and holdout variants only |
| Claim Accuracy Rate | Gold Stopped / (Gold Stopped + Gold Missed). See Claim Finding. |
| Detection-Only Rate | Missed with detected: true / bad canary total |
| Path Validity | Good canaries delivered and passed / good canary total |
| Black Hole Rate | Canaries voided by black holes / canaries released; and black hole time / observed time |
| Witness Accuracy | Proof Witness agreement with committed canary effects, per run |
The rules
- Efficacy Rate is always published with False Block Rate.
- The Blind and Void shares are published beside every rate they would otherwise inflate (see Absence Ladder).
- Rates are reported at the attribution level the proof supports: per SUT, per DUT, and per CUT only where differential attribution establishes it. Never finer.
- Detection without prevention is not efficacy. It shows up in the Detection-Only Rate.
- Rates from runs that are only Executed, not registered, are labeled unregistered. See Conformance Level.