How to verify a football prediction site — including this one
The predictions niche is full of leaderboards showing «ROI 550%» and «83% win rate» — numbers that are mathematically impossible over real samples. You don't need to trust anyone's marketing, including ours. Here are five checks that work on any prediction service. We apply all five to ourselves, publicly.
Last updated: October 1, 2026 — this guide changes when our own verification practices change.
1. Timestamped tips
A tip only counts if you know exactly when it was published — before the match, not after the result.
How we pass it
Every prediction is published before kick-off and settled from the same ledger that powers our accuracy blocks. Finished matches stay on their pages with the pre-match prediction next to the final score.
How to check any site
Ask: can you see WHEN each tip was posted? If a service only shows results without timestamps, the record can be assembled after the fact.
2. Full record — wins AND losses
A win rate without losses shown is a filtered record. Screenshots of winning bets prove nothing — anyone can edit a screenshot.
How we pass it
Our public record shows settled tips with losses, including the longest losing run — on purpose. Nothing is removed after a bad week.
How to check any site
Red flags: only big wins displayed, no losing runs anywhere, «last 30 days» windows that reset. Ask for the complete history, not highlights.
3. Sample size
Below ~100 settled picks, any win rate is noise. A 60% strike rate over 40 picks and the same rate over 400 picks are completely different claims.
How we pass it
Our record table shows the settled count per market next to every rate — Corners has a smaller sample than Match Result, and we say to read them differently.
How to check any site
Do the math yourself: hits ÷ settled tips. If the site won't show the denominator, the numerator is decoration.
4. Beating the closing line
The closing odds are the market's sharpest estimate. A service whose tips consistently get worse odds than the close has no real edge — whatever the win rate says.
How we pass it
Match pages show the odds drift since our tip (e.g. «2.30 → 2.05 ▼11%») — the market moving towards our prediction is the honest signal of value.
How to check any site
If a site never shows line movement, it cannot demonstrate value. Ask how their published odds compare to closing odds over the last 100 tips.
5. No editing, no deleting
Past predictions must be immutable. A service that can edit or delete losing tips after the fact has no record at all.
How we pass it
Two distinct phases, and we treat them honestly. Before kick-off a prediction can be regenerated when the model itself improves — a new statistical factor added, weights recalibrated, probabilities re-fitted. That is the model updating its opinion, and every page shows its «Updated» timestamp. After the final whistle the prediction is frozen and settled against the result — that record is append-only and is never edited or removed.
How to check any site
Test it: find an old losing tip on the service, check weeks later if it still exists with the same numbers. Changed odds, changed picks, vanished posts — walk away.
What passing all five does — and doesn't mean
Passing these checks means the service is honest, not that you will profit. Even a genuinely profitable model hits losing runs — with a 55% strike rate, a run of five losses is a normal statistical event. Judge any record over hundreds of picks, stake flat (1–2% of your bankroll per bet), and never bet money you can't afford to lose.
One nuance we apply to ourselves: a prediction published before a match may be regenerated as the model improves (new factors, recalibrated weights) — the «Updated» timestamp on each page shows when. What is never regenerated is the settled record: once the match is played, the prediction shown against the result is the one that was live before kick-off, kept unchanged forever.
See it applied: Our full record · How we predict · Responsible gambling