One scoring engine for every forecaster
CERES is scored against the simplest honest rival, persistence, and every forecaster is held to the same targets. A forecast earns credit only for what it adds.
The baseline to beat
Persistence copies the last observed food security phase forward and calls it the forecast. It is often hard to beat, because most situations do not change in 90 days. A system that cannot do better than persistence has not earned the trust of anyone planning aid.
The same rules for everyone
One scoring engine grades every contender against the same targets. CERES is not allowed a friendlier yardstick than the baseline, and the grades are written to a public record that cannot be rewritten.
Each forecast becomes gradeable 90 days after it is made, and grading against the observed outcomes began in September 2026.
