PredictMargin turns scheduled football matches into clearly explained home-win, draw and away-win probabilities. The product is built around a simple promise: publish before kickoff, preserve what was published and score the model after the result.
An analytics product, not a tip sheet
We publish probability distributions and likely scorelines so readers can understand uncertainty. We do not publish bookmaker odds, affiliate offers, value-bet labels, stake sizes, sure wins or guaranteed returns.
A record that can be checked
Every eligible forecast stores its generation time, training cutoff, model version and kickoff deadline. Completed matches keep the original pre-match forecast alongside the result, so poor forecasts cannot quietly disappear from the record.
Quality before page count
A fixture is withheld from search indexing when its feed is stale, its training sample is too small or the prediction is a demo or fallback. Languages and archive pages are expanded only after they have enough distinct, verified information to be useful.
Transparent limitations
Football is noisy. A 60% outcome still fails often, and a likely score is only one point in a distribution. Current models do not claim to know late injuries, confirmed lineups or every match-day condition unless a verified source supplies them.
Football data and attribution
The active data source is identified on published pages. Provider data remains subject to its own terms and licence; PredictMargin's model output and editorial explanation do not imply provider endorsement.
- Data verified
PredictMargin is not affiliated with, sponsored by or endorsed by any league, club, federation, data provider or bookmaker.