The forecast scorecard
A prospective record of forecasts Akashic actually published before the election — preserved as released, then scored against the certified result.
What will be scored
Each edition below was locked and timestamped before the result. When certified 2026 outcomes arrive, every call is scored automatically and the original forecast remains preserved exactly as published.
Four things will be reported: the Brier score on win probabilities, the calibration curve, the directional record with every miss named, and the seat-forecast error. If two editions are scored, the comparison will show whether the revisions improved the call.
Why there is no earlier score
Akashic did not exist as a public forecaster in 2018 or 2022. Re-running a model on known history can be useful research, but it is not the same as publishing a forecast before an election. This page starts where the real record starts: 2026.
The forecast method and scoring rules are documented on the methodology page. Retrospective model research may inform calibration, but it is never presented as an Akashic forecast or part of this track record. · Back to the forecast →