Glossary
| Term | Meaning in Supercast |
|---|---|
| As-of time | The declared timestamp of a forecast revision |
| Base rate | An outside-view frequency or predictive probability from comparable cases |
| Brier score | Squared probability error under the stated binary or categorical convention |
| Calibration | Agreement between predicted probabilities and outcome frequencies |
| Cluster | A group of related questions kept together for splitting or resampling |
| Coarsened Brier | Score after replacing each forecast by its bin’s mean probability |
| Conditional likelihood | Probability of evidence under a hypothesis and earlier evidence |
| CRPS | Continuous ranked probability score; an outcome-unit measure of predictive distribution error |
| Cutoff | Latest allowed evidence or outcome availability time for an analysis |
| Degenerate bootstrap | A resampled distribution with no observed variation |
| Evidence origin | Identity of the underlying observation, independent of its republications |
| Holdout | Data excluded from selection and used to evaluate the selected method |
| Idempotency key | Request ID that identifies an exact retry of a durable mutation |
| Information gain | Expected entropy reduction from observing an indicator |
| Logit | Log odds, defined for probabilities strictly inside zero and one |
| Proper score | A rule whose expected optimum is honest distribution reporting |
| Reference class | Past cases selected using a stated comparability and inclusion rule |
| Resolution | An adjudicated outcome, distinct from the discriminating ability also called resolution in calibration analysis |
| Sensitivity envelope | Output range from alternative stated assumptions, not a statistical confidence interval |
| Sharpness | Concentration of predictive probabilities or distributions, independent of outcomes |
| Unresolved | The required outcome is not established; never an automatic no |
| Value of information | Improvement in expected decision utility from observing a signal, optionally less cost |
| Void | An administrative exclusion with an explicit reason |
When a term has more than one meaning, report the quantity and convention rather than relying on the label alone. “Resolution” is the most important example: journal adjudication and forecast discrimination are separate concepts.