How the predictions work

Know the data, the model and the limits behind every number.

Method

Every fixture starts from the same data: results, team statistics, line-ups, absences, referees and bookmaker prices supplied by Sportmonks. A statistical model estimates how many goals each side is likely to score, and the forecasts for every market are read from that single picture so they stay consistent with each other.

The model's view is pooled with the provider's prediction and the bookmakers' prices, weighted by how well each has done historically. Forecasts are recorded before kick-off and graded afterwards on the Results page, so the track record is public.

Probabilities describe uncertainty. They are not betting advice; please read our prediction disclaimer and responsible gambling page.

From results to probabilities

  1. Start with completed results. The model uses completed matches that kicked off more than three hours before the fixture, which keeps same-day information out.
  2. Estimate each team's scoring rate. Recent scoring and conceding records are blended toward league averages to temper small samples, and home advantage is included.
  3. Build a score distribution. The two goal rates give a probability for every scoreline, and from the scorelines come home, draw and away, goal totals and the other markets.
  4. Show the evidence. Each analysis shows how many matches it used; limited history can make a forecast unavailable or less informative.

What the model cannot tell you

There is no guarantee of accuracy. The forecasts are tested on held-out past matches and the results are published, but no independent party has audited them. A 70% estimate means an outcome is likely, not that it will happen.

Transfers and tactics are not model inputs, and forecasts stay pre-match estimates: a score update during a match does not make them a live forecast. A likely outcome alone says nothing about the value of bookmaker odds.