Straight answers,
no small print.

Every fixture is assembled from one licensed data supplier and reduced to a brief: recent form for both sides, their head-to-head record, home and away splits, league position and scoring rate, what each side plays next, confirmed line-ups where they exist, and the price every bookmaker quoted with its margin stripped out. An AI model reads that brief and makes three calls — the most likely score, the safest bet, and the best bet, meaning the one where its judgement differs most from the bookmakers. The three are checked against each other before publication, so a predicted 1-1 can never appear beside an over-2.5 pick. Every settled pick is then graded against the real result and published, win or lose.

It is a rating from 1 to 10, and it ranks our own picks against each other rather than claiming a probability: a 9 is a pick we hold more strongly than a 6, not a claim that it lands nine times in ten. We say that plainly because the difference matters — a number that looks like a percentage invites you to price it like one. What the rating can be held to is order: our 9s should win more often than our 6s, and if they do not, the rating is not doing its job. Every settled pick is graded in public, win or lose, so the record behind the rating is there to be read. Note too that the strongest rating is not the best bet — the safest pick and the one carrying real value are different questions, and we publish both. Across 4,095 graded picks the published rate is 68%.

Predictions run on Football today. Basketball, Hockey, Baseball, American Football, Cricket carry live scores, fixtures, results, tables and season statistics, but the prediction model for them is being rebuilt rather than left running in a state we would not stand behind. When it covers a sport, that sport's predictions appear — and are graded on the same public record as everything else.

No, and anyone telling you otherwise is selling something. A prediction is a judgement, not a promise. Matches are decided by red cards, deflections and refereeing calls that no model can price, and our highest-rated picks still lose — if they never did, the rating would be a lie. Use the picks as one input, weigh the rating on them, and make your own call.

A fixture is predicted as it comes inside the two-day window before kick-off, so the pick is made on the most complete team and form data available rather than on a guess made a week out. Once published it is frozen. Nothing is rewritten after a result comes in, because a record you can edit is not a record.

7 markets: Both Teams to Score, Over/Under 2.5, Correct Score, Sure Banker, Sure Straight Win, Double Chance, Asian Handicap. Not every market applies to every sport — a market that cannot be settled is not published, so you will never see a both-teams-to-score prediction on a basketball fixture, where both sides always score. Each market also has its own page listing that day's picks across every fixture.

No. The same process runs on every fixture in the feed, from the Champions League down to second and third tiers most tipping sites ignore. The market pages pull from all of it, not just the top five leagues.

When a match finishes we check the pick against the real result and keep it for good, right or wrong. That record currently stands at 2,800 correct from 4,095 picks checked, broken down by sport and by how sure we said we were. Losing runs stay in the total. Picks we cannot check honestly — a postponed match, a bet that was void — are left out of the count completely. They are never counted as wins.