SARDASPORTS: football predictions, every one gradedSARDASPORTS

No. Every board and the full record are free to read without one. A free account (18+, confirmed by e-mail) is where premium lives: early access to the editor's picks and members-only write-ups. Premium never changes the engine's predictions or the record.

Yes. The track record page lists every graded pick, wins and losses, and the full history can be downloaded as a spreadsheet. Nothing is removed from the record.

Fixtures, prices and results are pulled in on a schedule through the day. A fixture appears on the board with its picks once it has been ingested, and its picks can change as new results arrive before kick-off. The board always shows the latest view.

The first three show the home win, draw and away win prices. The 1X2, O/U and BTTS columns show the engine's pick in each market with its price. Best tip is the one pick the fixture leads with, and the ring is the confidence score out of ten. A price marked ≈ is the engine's own fair price, used where no bookmaker price was stored.

It ranks how closely the engine's three estimators agree on the fixture against every fixture it has priced, out of ten: 5 is typical and 9 or more is in the top tenth. It measures agreement, not the chance of winning; the percentage beside each pick is that. Picks the estimators agree on most have won more often than those they disagree on, but only modestly, which is why bankers are chosen on probability instead.

From the markets priced for the fixture, the engine takes the likeliest pick among those that beat how often that market lands anyway, measured in the fixture's own competition: over 2.5 goals is routine in one league and rare in another, so a line is only a finding if it beats its own league's rate. Draw no bet is one of the markets: back a side and the stake is returned on a draw, which is why a draw is left out of the win rate. Corners and cards are priced and listed on the match page but are never the best tip. Where nothing qualifies, the likeliest pick overall is shown.

A banker is a pick the engine rates at least 80% likely. The Bankers page lists them for each day, with the record of how the settled ones went. A near-certain pick usually carries a short price, so a banker is a safe pick, not necessarily a valuable one.

Not any single match, and no model can. The engine produces probabilities and is judged on whether they hold up over many matches. A pick rated 70% is expected to miss three times in ten.

43 competitions across 32 countries, from top domestic leagues down to divisions most sites skip. The leagues page lists them all, with how much history sits behind each.

Yes, on their own pages. Basketball covers the NBA, the EuroLeague and dozens of other leagues, with game winners and total points; tennis covers every ATP and WTA match, with match winners and total games. Their predictions are stored before the match and graded in a record of their own, kept apart from football's. They are younger than the football record, so the samples are small, and each page shows how every model measures against the simple rule it has to beat, including where it does not.

No. A win rate says how often picks land, not whether the price paid for them was good. A short price on a near-certainty is usually priced about right already, so winning often can still lose money.

No to both. The site is for information, prices are shown for context, and nothing here links to a bookmaker. If you do bet, stake only what you can afford to lose. You must be 18 or over.

No. Every pick is a statistical estimate produced by the engine. Nobody can sell a guaranteed result, and this site does not claim to.