RadarFooty
Account
Not “who wins?”. What the record shows, what the context says and what an honest model estimates, with its own track record published. Data, context, explanation. You decide the rest.
One page with what we follow before the matches: your teams, the matches where something happened more often than in the rest of the league, the streaks still alive, the title race and the round, before and after. It shows what happened, not what will happen.
The chances are our model's estimate for each result (1 = home win, X = draw, 2 = away win), not advice. They are frozen before kick-off and you can check them afterwards: the record is public.
Track record of frozen estimates: in 97 played matches the estimate's favourite won 56 (58%; 95% interval: 48%–67%). Brier score: 0.577, against 0.635 for the league frequencies (lower is better). The estimates ledger →
Football statistics and soccer stats for Europe's top leagues: team stats, results and fixtures, head-to-head records, Elo ratings, match analysis with context, a season simulator and calibrated match probabilities whose record you can check.
What we do differently. Every number is measured, with how many matches it comes from and the league next to it. Every model estimate is frozen before kick-off and stays published with the result: the estimates ledger.
A site Ambassador publishes picks on matches, with our model's probabilities and the Arena's points. Anyone can see the page; results are checked automatically after the final whistle.
See the picks →Each bar is one week: what share of the bots' picks came true. The pale bars are reconstructed weeks ↺, from before the bots picked in real time; the dotted line is what the model estimated for the same picks. The more matches in a combination, the less often it comes true — and the more points it earns.
real-time week reconstructed ↺ what the model estimated
The top bots in the last week with results: the points from the picks that came true, as in the Arena's weekly standings.
The players who chose to appear in the standings: the points from the picks that came true in the latest period with results, as in the Arena's community standings.
The most valuable picks that came true, from players who made their picks public: the less likely the combination was, the more points it earned. Only matches already played.
Ways into the data. Each card shows a real figure taken from the page it opens, updated with the site; open it to see where the figure comes from and what it does not say.
Form at home and away, streaks, head-to-head, half-time and full-time; goals by minute and late goals where the source has the minutes (today, Liga 1). Every number comes with how many matches it is based on. Nothing hand-typed, nothing invented.
Days of rest, fixture congestion, travel distance, weather at kick-off, the referee (with their record on the referees page), what the table position means. Shown in real units, side by side, home versus away.
Each estimate is frozen before kick-off and stays published, with the result, in the estimates ledger. The calibration page shows how often “60 %” actually happened in past seasons. If the model is off, you will see it here first.
Premier League, La Liga, Serie A, Bundesliga and Ligue 1 this season, with past seasons from the free source; Liga 1 as history. Open now: match, team and competition pages, the season simulator, fixture difficulty, the round on radar, match filters and active streaks, the estimates ledger, the daily game, the Arena and accounts (email and password, or Google).