Methodology

Every published metric on the Public Trust Centre traces back to a sealed Prediction Record field. These pages explain — in plain language — what each concept means, how it is derived, and what Chresma deliberately does not claim.

Evidence Quality
How Chresma describes the reliability of each piece of evidence.
2 sections
Evidence Sufficiency
Whether the retrieved evidence is enough to support a forecast.
2 sections
Confidence
How certain Chresma is in a committed forecast, as opposed to the forecast itself.
1 section
Probability
The estimated likelihood of the primary outcome, when Chresma commits to one.
1 section
Calibration
How closely stated confidence matches observed reliability.
1 section
Brier Scoring
A proper scoring rule for probabilistic forecasts.
1 section
Abstention
Chresma refuses to forecast when evidence is insufficient or the question is unanswerable.
1 section
Contradiction Analysis
How Chresma quantifies conflict between retrieved evidence.
1 section
Outcome Resolution
How Chresma compares a prediction to the eventual real-world outcome.
1 section
Deterministic Replay
Every sealed record can be re-derived from its sealed inputs.
1 section
Record Sealing
How a prediction becomes an immutable, verifiable record.
2 sections