The model and the market
The goals model uses a Dixon–Coles scoreline distribution. It describes possible scores and supplies estimates for related markets, including combinations whose outcomes are correlated.
The selection system also uses bookmaker market probabilities with the margin removed. Model and market contributions can differ by market and configuration. A model estimate is not an independently proven edge simply because it differs from a price.
Forecast, watch item or pick?
A forecast is an estimate of what could happen. A watch item invites closer inspection. A published selection is the outcome chosen by the applicable selection rules; those rules include fallback tiers, so a headline alone must not be read as proof of a positive-value qualification.
The Main Outcome identifies the fixture’s headline selection. Conviction is a display summary, not a promise of success. Inspect the probability, available price, uncertainty and recorded tier together.
Keep the record with the claim
Committed selections carry a selection time, and settled outcomes are kept on this site. The record should be read by product and date window, with wins, losses, prices and the relevant denominator in view.
A Builder combines outcomes from the same fixture, where correlation matters. An accumulator combines fixtures and increases variance. Their records and risks must remain separate from single selections.
Numbers before narrative
Explanations should describe the available numerical evidence and its limitations. Written analysis is not an additional probability estimate, and language alone cannot validate a model or a price.
Missing or stale information weakens a decision. When a source is unavailable, the product should say so rather than substitute a confident-looking number.