Editorial Policy

Analysis on fixture pages is generated from the underlying statistics, not written to justify a selection after the fact. Where commentary appears, it describes what the data shows and names the figure it is based on.

Predictions are timestamped and are not altered once a match has kicked off. When a prediction is regenerated before kickoff because new data arrived, the page shows the updated time.

Model performance is published on our accuracy page with hit rates, a calibration score and the sample size, including periods of poor performance. We do not cherry-pick winning runs.

Factual errors are corrected as soon as they are confirmed. Material corrections are noted on the affected page. Report an error through our contact page.

Odds displayed for comparison do not influence model output. Any commercial relationship that could affect what we publish will be disclosed clearly on the pages it affects.