Every football prediction, on the record

The full history of our football picks, open to anyone. We log each pick before the match starts, grade it won or lost once there is a full-time score, and never remove it afterwards. The daily squares, the month bars and the fortnight list all read from that single archive, so their numbers always match.

It starts on 18 May 2021 and ends on 2 Oct 2026, a span of 1,888 days in which 281,317 picks were settled. The heatmap shows the latest 26 weeks, and the monthly bars the latest 12 months.

We divide the picks that won by all picks that were settled: 187,117 of 281,317, or 67%. Each match counts once, whatever day it falls on, so a busy Saturday weighs more than a quiet Tuesday. Averaging the daily rates instead would give a light day as much say as a full one, and we avoid that.

It is one of the few matches each day that the model gives a chance of 90 percent or more. Of the 9,318 settled so far, 6,627 won, which is 71%, compared with 67% across all our picks. A shorter, stricter list should hit more often than the full one, and showing both figures side by side is how you can check that it does.

As soon as the match is over and the full-time score is in. Each fixture gets a single pick, posted before the line-ups are announced, and that score settles it as won or lost, with no half results. If a game is postponed or abandoned it never gets a full-time score, so we drop it from the count instead of logging it as a loss.

The newest graded date is 2 Oct 2026. Today is missing because its games are still in progress: a day is added only when all of its matches have finished and every pick on it is settled, and from then on it does not change. The heatmap, the month bars and the fortnight list are rebuilt together from one source, so none of them can get a day ahead of the others.

A spreadsheet covering the last twelve months, split into one tab per season with a row for every match: the teams, our pick before kickoff and the final result. It is free and needs no sign-up. Every percentage on this page is calculated from that file, so you can redo the maths yourself.

Each square covers a single day, and one day only holds a handful of settled picks. If four were settled, the square can show just five values, so one late goal moves it by 25 points. That swing is what small samples do, not a fault. For the trend, look at the month bars and the overall win rate; a single square on its own says very little.

All of them. Every competition the model rates is included, from Europe's top divisions and the tiers below them to domestic cups and continental tournaments. The same goes for each market that can carry a best tip: match result, both teams to score, double chance and total goals. We leave nothing out, whether a poor month, a league we misread or a market that went against us.

Yes, and we encourage it. When the final whistle goes, the pick becomes visible to all visitors, won or lost, and it remains exactly as it was posted. Losing picks are never removed. Each one still counts in its day, its month and the overall total, and that is why the win rate means something.

No, and we want to be clear about it. A win rate tells you how often picks were right, not what odds you got, and the odds matter just as much. Any model, even a good one, has losing runs. Read each pick as a view on what is likely rather than a certainty, bet only money you can do without, and contact BeGambleAware.org if betting no longer feels like fun.