Prediction Oracle API

A useful interface names the answer type

Prediction oracle API is a broad descriptive term, not one universal protocol. An implementation might return a model estimate, a source-backed answer or a resolved event result. The response should tell the client which one it received.

A conceptual envelope can include a stable question identifier, schema version, answer type, status, issue time, horizon and evidence or method references. The exact fields depend on the service; this site does not operate a live prediction endpoint.

State the event and the horizon

A probability is incomplete without a defined event and time window. Specify whether the estimate concerns occurrence during a period, occurrence by a deadline or a state at an instant. Preserve the wording and rule version used to make the prediction.

Keep issue time separate from the forecast horizon and data cutoff. A revised estimate should not erase the earlier record. Evaluation requires knowing what was predicted before the outcome became known.

Handle uncertainty and nonanswers honestly

Use explicit states for insufficient evidence, unsupported questions, pending resolution and temporary service failure. Do not turn missing probability into zero or an unresolved question into false. An explanation is not a substitute for evidence.

Forecast quality can be evaluated against resolved outcomes and suitable baselines. The forecast verification note identifies probability-verification background. No single sample or attractive metric establishes reliable performance across all question types.

Explore a local response example

The Oracle API guide includes downloadable static JSON fixtures for a forecast, an observation and an attestation. They illustrate interface design only; no request is sent to an external provider.

For settlement rather than estimation, continue to Prediction Market Oracle. For the wider idea of a queryable source of answers, read Prediction Oracles.