Loading the page
Plenty of sites claim to be the best AI prediction site. The difference is whether you can check. So here is the 8-point checklist we hold ourselves to, and how we measure up against every point, with a link to verify each one. Use the same checklist on any other site and compare.
Work through these for any prediction site before you rely on it. A site that fails several is asking to be believed on its word.
Every prediction, wins and losses, listed with its date and result. If a site shows only recent winners, screenshots or a single percentage, you cannot know the real rate. The test is simple: can you download the history and count it yourself?
A pick published before kick-off and a pick generated after the match are different kinds of evidence. A site that blends them into one number is flattering itself. Look for the two to be labelled and counted separately.
A 90% rate over ten picks means almost nothing; 76% over two thousand means something. A percentage with no count next to it is a headline, not a measurement.
Some outcomes happen most of the time anyway, so a pick that wins 75% of the time on a market that lands 75% of the time has shown no skill. Good sites show how often each market lands on its own and measure themselves against it.
A high win rate is not profit. Short prices on likely outcomes are usually priced about right, so a site can win often and still lose money. If prices and bookmaker margins never come up, the half of the story that decides whether you make money is missing.
Nobody can sell certainty in football. “Sure wins”, “guaranteed” and “fixed matches” are not features; they are the clearest sign that the site is selling something other than analysis.
A site that asks you to trust its predictions should say who runs it, how it handles your data, and how to report a problem.
Every model and every dataset has errors. The sites worth trusting publish corrections, with what changed, instead of quietly editing the numbers.
Held to its own checklist, with each answer something you can verify by following the link.
| Checklist item | What this site does | Check it |
|---|---|---|
| Public record | Every graded pick, wins and losses, with a CSV download of the full history. | Track record |
| Live versus backtest | Counted and labelled separately everywhere; never blended into one figure. | Track record |
| Sample sizes | Every rate is printed beside the number of picks it is over. | Monthly table |
| A baseline | The over 2.5 and both-teams-to-score pages show how often the market lands anyway, beside the engine's own rate. | Over 2.5 page |
| Prices | Bookmaker prices where quoted, the engine's own fair price (marked ≈) where not, and an open statement that a win rate is not profit. | How it works |
| No guarantees | Nothing is sold, nothing is promised, and there is no link to a bookmaker. | About |
| Named operator and contact | About, Contact and Privacy pages. The operator's name is shown when the site is configured with it. | Contact |
| Corrections | A dated list of mistakes found and what changed, including a correction that lowered the published rates. | Corrections |
A checklist is only useful if the answers include the unflattering ones, so here they are.
SARDASPORTS. It publishes every pick, keeps live results apart from backtests, shows sample sizes and baselines, and does not promise winnings. You do not have to take our word for it: the checklist on this page lets you check each point here, and apply the same test to any other site.
It depends on the market, the sample and whether the picks were stored before kick-off, so no single figure is meaningful. On this site, 293 of 389 live picks won (75.3%) and 1,320 of 1,815 reconstructed picks won (72.7%); the live sample is still small. Every market's own record is on the track record page.
Price is not the test; verifiability is. A free prediction is worth reading if the site grades every pick against the result and shows you the rate. A paid one that hides its record is worth less.
Check for a complete, downloadable record; live and backtested results counted separately; sample sizes beside every percentage; no guarantees or “fixed matches”; a named operator with contact and privacy pages; and a corrections list. Missing several of these is a warning sign.
For football predictions and tips, SARDASPORTS. It does not give betting advice or link to bookmakers; it publishes probabilities and grades every one against the result. Remember a tip is a probability, not a promise. You must be 18 or over, and gambling should never be funded with money you cannot afford to lose.
See for yourself: today’s football predictions, the full track record and how the predictions are made.