What we forecast, what the ratings mean, and how every call is graded.
MatchdayAI produces a fresh analysis for every fixture close to kickoff, built from publicly available football data. Each analysis covers several outcome types and is paired with a plain-English narrative and a confidence rating. Once the match settles, the forecast is auto-graded against the official result.
This page describes what we publish and how it is scored. The models themselves, the inputs they weigh and how those inputs are combined are not published — that is the part of the product we build, and describing it in detail would tell you nothing you can check anyway. What you can check is the record, which is why every forecast is graded in the open.
For each fixture we evaluate several outcomes independently:
Each forecast is labelled low, medium, or high confidence. Confidence is not a guarantee — it is relative model conviction, and it is only worth what the record says it is worth, so we publish how well each tier has actually performed and let you judge the rating itself.
Each forecast comes with a written case for it, citing the statistics behind the call. Logged-in users see the full analysis; logged-out users see a short preview.
When a fixture finishes, every forecast is scored automatically against the official result — hits and misses alike. Aggregate accuracy is published openly on our Track Record page, broken down by outcome type, confidence tier, and competition. Nothing is retroactively edited or removed.
Our outputs are probabilistic models, never certainties. MatchdayAI is an information and analytics service: we do not accept bets, process wagers, or hold customer funds, and nothing here is financial advice or a recommendation to stake money. Past model performance does not guarantee future results. See our terms for the full position.
