Our football prediction track record

A public scoreboard for the football tips we publish. Every pick is written down before kickoff, marked won or lost against the full-time score, and then left there for good. The squares, the monthly bars and the two-week list are three views of one archive, which is why they always agree with each other.

1,885 days of it, opening on 18 May 2021 and running to 29 Sept 2026, with 280,437 tips settled in between. The squares cover the past 26 weeks and the bars the past 12 months.

Tips that came in, over tips that were settled: 186,545 out of 280,437, which is 67%. The sum is done across matches rather than across days, so a packed Saturday pulls more weight than a thin Tuesday, and the headline is a true average instead of an average of averages.

A high-confidence tip sits in the model’s top tier, the handful of fixtures a day it scores 90 or more out of 100. 9,296 have been settled and 6,610 came in, 71%, against 67% for everything we put out. A tighter list ought to win more often than a loose one, and the only way to see whether it does is to print the two apart.

The moment its match has a full-time score. One tip goes out per fixture, published before the teams are out, and the result decides it. Every tip is either won or lost — nothing in between. A game called off or abandoned never reaches full time, so it leaves the count altogether rather than going down as a defeat.

The last day on the record is 29 Sept 2026. Today’s fixtures are still being played, so they are not on it yet: a date arrives only once every match on it has a result and every tip has been marked. After that it is fixed. The squares, the bars and the two-week list are all rebuilt from the same file, so no panel can run a day ahead of another.

One spreadsheet, a sheet per season, a line per match: who played, what we said beforehand and how it finished. Twelve months of it, free, and it asks for no account. It is the source these percentages are added up from, so you are welcome to add them up again.

Because one square is one day, and a day holds very few matches. Where four were settled, that square has only five shades available to it, and a late goal is worth 25 points of win rate on its own. Nothing has gone wrong when it swings; that is simply what a short denominator does. The bars and the all-time figure carry the signal, and any single square is close to noise.

Everything the model rates: the biggest divisions in Europe, second tiers below them, national cups and continental football. And every market we name a best tip in, which means match result, double chance, over and under goals, and both teams to score. Nothing is held back from the count, not a weak month, not a league we read badly, not a market that let us down.

Please do. When a match ends, its pick opens to everyone, winner or loser, and it stays on the site exactly as it was published. A losing tip is not quietly taken down. It keeps its place in the day, in the month and in the total, which is the only thing that makes a number like this worth reading.

Not on its own, and we would rather say so plainly. A percentage only counts how often tips came in: the price you take on each bet matters just as much, and even a good model has bad weeks. Treat a tip as an opinion about what is likely, never as a certainty, stake only what you can afford to lose, and talk to BeGambleAware.org if betting has stopped being fun.