Model ratings

Every team gets two numbers. Attack is how many goals it scores relative to the average side in its division, and defence is how many it concedes on the same scale, so 1.00 is exactly average and for defence lower is better. An attack of 1.25 means a quarter more goals than a typical side in that league. They are fitted from this season's results with last season as a prior, and recent matches count for more.

xpts turns that pair into one figure: the points per game the model expects this team to take against an average side on neutral ground. It is read off the same scoreline grid the match cards use, which is why a team that tops this table can never appear as an underdog on its own card. For a side rated exactly average the arithmetic collapses to 1.5 − half the draw chance, so an average team scores around 1.37 — and no average team can ever reach 1.5.

pl is how many league matches this season sit behind the rating, and it is the column to read first. Under 6 matches a rating still rests mostly on last season, and those rows are dimmed to say so. A promoted side has no record in the division at all, so it starts from the promoted default — hover its name to see that noted. Hover any figure for what it means in goals.

The full derivation — the recency weighting, how last season enters, the iterative solve, and every parameter above — is on the method page. The ratings that feed the tables above are also the ones behind this week's fixtures.