Predicted scorelines and win probabilities for football's biggest matches — powered by expected goals, current form and head-to-head record. Clear, honest, and built to compare.
★ Match of the day
The model makes Germany 48% to win, the draw 26% and Netherlands 26% — fair odds 2.08 · 3.85 · 3.85 — with 2-1 the most likely scoreline, at medium confidence. Edge to Germany: the better side on the model's ratings, but not by much. In past matches rated this way the favourite won 54%, 26% were drawn and the underdog won 21%.
The day's most important matches across every competition we cover — ranked by our model. See all matches — recent results and the days ahead — on the fixtures page.
The 13 most important matches on 16 November 2026, across every competition we cover, ranked by the model.
6 of the most important matches played on 15 November 2026, with the final score and the call the model published before kick-off. Across all 22 it called 10 correct — 45%, against 49% over every prediction on record.
Previews and reviews built from the model's own probabilities and expected goals — what the numbers say about the biggest matches of the next seven days.
The model makes it 41% Denmark, 26% draw, 33% Norway — most likely 2-1 (fair odds 2.44 on Denmark).
17 November 2026 · UEFA Nations LeagueThe model makes it 51% Liverpool, 23% draw, 26% Manchester United — most likely 2-1 (fair odds 1.96 on Liverpool).
22 November 2026 · Premier LeagueThe model makes it 62% Barcelona, 21% draw, 17% Villarreal — most likely 2-1 (fair odds 1.61 on Barcelona).
22 November 2026 · La LigaThe model makes it 42% Chelsea, 27% draw, 31% Leeds — most likely 2-1 (fair odds 2.38 on Chelsea).
21 November 2026 · Premier LeagueThe model makes it 46% AC Milan, 25% draw, 29% Frosinone — most likely 2-1 (fair odds 2.17 on AC Milan).
22 November 2026 · Serie AThe model makes it 22% Atalanta, 24% draw, 54% Inter — most likely 1-2 (fair odds 1.85 on Inter).
22 November 2026 · Serie AEvery prediction is logged before kickoff and checked against the real result — a transparent, automatically-updated record of how the model actually does.
The hit rate is only half the question. The full track record scores every settled prediction against an uninformed baseline and shows, band by band, whether a 70% from this model really happens 70 times in 100 — including where it does not.