A plain-English look at the reasoning behind every number on this site โ and where we're honest about its limits.
There are two models, fitted separately, because the two competitions behave differently.
Clubs. Each league is fitted on its own results lake: every club gets an attack and a defense rating, recent matches counting for more. Club fixtures are priced with the home advantage the league’s own results imply โ a real ground, a real crowd. Where a league or a promoted club has too little history to fit, that fixture falls back to the market-anchored price and says so on the match page.
Nations. Each nation gets the same pair of ratings from its full competitive history, with lopsided friendlies counting for much less. World Cup ties were the exception on home advantage: matches at a neutral venue got none, and only host-nation matches at a real home ground did.
Bare results lag reality, so we layer in two things a scoreline alone misses: how good each side's recent chance quality has actually been (shots and expected goals, not just the final score), and โ for national teams, whose rosters turn over between tournaments โ the club-level quality of the players in the squad. Together these sharpen the ratings beyond what the win-loss record shows.
Those ratings produce a full picture of how a given match is likely to unfold โ which feeds every market on the site in one internally-consistent shot: match result, Over/Under, Both Teams To Score, handicaps, corners, cards, scorers and more.
A match's expected goals are distributed across the players likely to be on the pitch, weighted by position, by each player's own recent shot and goal rate, and by whether they are expected to start. Every player's rate sums back to the team total by construction, so a scorer projection can never disagree with the match projection it came from. Anytime-scorer probabilities are then calibrated against how our own graded projections have actually landed.
For a knockout tournament we play the whole thing out tens of thousands of times โ all 12 groups, the best third-placed teams, and the full knockout bracket โ to produce advance, win-group, reach-final and title odds. A measure of uncertainty is added to every run because the new 48-team format is genuinely high-variance.
A probability is only meaningful next to the price. Across the site we show our number alongside the best available market price so you can see exactly where the two disagree โ and decide for yourself. We don't sell picks; we show our work on the outcomes and let the comparison speak.