When we say 75%, it happens about 75% of the time

Loads of tipsters throw around confidence numbers. We do something different: we save every pick before the match kicks off, then check whether it came true. It turns out our percentages are almost bang on, and here's the proof.

97%
Our percentages are this accurate
When we say a pick is likely, it lands about as often as we said. On average we're within a couple of points.
72%
Overall hit rate
How often our headline pick actually won, across every match we've tracked.
1,070
graded predictions in the calibration set. The public record counts the published picks only.
Every one saved before kick-off, then checked against the real result. Wins and losses.

What we said vs what actually happened

Each row groups our picks by how confident we were. The green bar is what really happened. The closer it is to what we said, the more you can trust the number. Simple as that.

"93%"
We said 93% It happened 88% Spot on from 26 picks
"83%"
We said 83% It happened 81% Spot on from 216 picks
"76%"
We said 76% It happened 79% Spot on from 422 picks
"65%"
We said 65% It happened 64% Spot on from 248 picks
"57%"
We said 57% It happened 54% Spot on from 137 picks
"45%"
We said 45% It happened 33% A touch under from 15 picks
"33%"
We said 33% It happened 50% Even better than we said from 6 picks

A few rows with only a handful of picks bounce around, that's normal with small numbers. The rows with hundreds of picks are the ones that matter, and those are almost exactly on the mark.

Why this matters to you

It's easy for a tips site to say "80% chance" and hope you don't check. If their numbers are inflated, you back things that aren't as safe as they look. Ours have been checked against 1,070 real results, and they hold up, so when you see a percentage on this site, you can take it at face value.

  1. We save it before kick-off. Every pick and its percentage is written down before the match starts, so we can't change our story afterwards.
  2. We check the real result. Win or lose, every pick is checked against what actually happened. Nothing is quietly deleted.
  3. We compare said vs happened. We group picks by how confident we were and see how often they came true. That's the chart above.

For the data-minded: this is a calibration check, and our Brier score is 0.190 (lower is better, 0 is perfect).

It's information, not betting advice, and nothing is guaranteed. But the numbers are open and you can check every one on the full track record. 18+ · gamble responsibly.

Pro

The singles we'd pick ourselves, with the research and the price behind each one

  • The daily singles. Each one cleared three checks: the line landed in at least 9 of the two sides' last 12 games, a UK price of 1.35 or more, and our model's number clears what the price needs. Usually 0 to 5 a day, one unit each.
  • The case for each, put together. The last 12 against what the price needs, who is missing, the head to head, and the best price at a UK bookmaker, in one place.
  • The desk's card on every fixture it covers. Its verdict, every market graded, and the reason written down. And tomorrow's singles the night before.

Every pick published before kick-off, graded in public. The losing ones stay on the page.

Seven days for £1 then £9.99 a month, or £79.99 a year. Cancel any time.
The pick of the day is free, in full, on days with two or more singles. 18+. BeGambleAware.org. Analysis and opinion, not betting advice.