A model estimate needs a separate test

An observed percentage describes finished matches in a defined sample. A model probability describes outcomes under its assumptions. A model can assign a number to an outcome without demonstrating that this number is calibrated against future results.

MatchDaySync's free club predictions are labelled a prototype, not yet validated. The model uses stored season scoring and conceding rates, scaled against the league's home and away scoring averages. It does not use player lineups, injuries, rest, weather or shot-level quality.

The current club prototype requires at least 5 finished season matches for each club and 40 finished league matches. It also needs a positive league scoring average. These are input gates; they are not a proof of reliable forecasts.

Figures are offered only for eligible fixtures whose known kickoff has not started. Missing kickoff information and insufficient samples remain unavailable. The restored snapshot is not a live source of current fixture or squad information.

Restored from the 11 October backup. Current updates are not connected.

The model uses independent Poisson goal distributions. The displayed score grid covers zero through seven goals per side and is normalised within that grid. Higher-score tail outcomes are outside the displayed grid, so the presented probabilities are not an uncapped full-score distribution.

The simulation draws repeated scorelines from the model. Its results can change between runs. Repeating a simulation explores the same assumptions; it does not add actual matches or validate the forecast.

The most likely individual score can still have a low probability because the remaining probability is spread across many other scores.

The World Cup 2026 archive preserves a different projection model and historical review record. Its calibration and archived confidence labels do not establish performance for today's club prototype.

Read the full club prototype methodology for its calculation and limitations, the World Cup archive methodology for that separate model, and the archived accuracy ledger for the historical record. Open the prototype prediction pages.

• The sample behind the rate: Denominators, valid zeros and missing values.
• Choose the window that answers your question: Season and recent-match windows, rounds, dates and half-by-half tables.
• Learn hub: Short guides to reading samples, windows and model estimates.

• Threshold explorer: Observed goal, corner and card frequencies in a chosen sample.
• Stats Studio: Compare selected club metrics with the measured matches shown for each.
• Form guide: Club results over stored last-5, last-10 or last-20 windows.
• Leagues: Tables, fixtures, results and statistics for each stored competition.

• Glossary: Short definitions of the statistical terms used across the site.
• FAQ: Short answers about projections, past purchases and our record.
• Data status: Coverage of the stored match data, competition by competition.
• CSV reference: Paths, columns and limits of the three public CSV downloads.