Every prediction on this site comes from one documented procedure that runs once a night. It turns bookmaker prices into fair probabilities, checks them against our own team-strength model, and publishes one main prediction per fixture — the most likely market that passes a set of logic checks, with limits that stop a single bet type from filling the whole day.
Two rules sit above everything else. First: we never invent a number. Where a fixture has no bookmaker price, no percentage is shown and the row reads no price. Second: picks are set once, overnight, and are not changed during the day. A pick, the odds it was published at and its percentage are a record, not a moving target. The only later addition is a price for a pick that had none when it was published, once a bookmaker opens a market for it — the pick itself stays as it was.
Fixtures, results, line-ups, injuries, live scores and bookmaker odds are licensed from commercial sports-data providers under paid agreements. We do not scrape third-party sites and we never re-publish another tipster's selections as our own.
The provider also supplies its own prediction for many fixtures — a short advice string and a set of percentages. That advice is one input among several, and it is used as the published pick only where a fixture has neither prices nor enough match history for our own model.
A bookmaker's odds include a margin, so the raw prices of every outcome add up to more than 100%. For each bookmaker separately we remove that margin, then average the fair probabilities across bookmakers when at least three price the market, leaving out any that sit far from the rest. For the three-way match result we use the Shin method, which corrects the tendency of simple margin removal to understate favourites.
Alongside the market we run our own team-strength model, refitted every night from recent results, with older matches counting for less. For match result, double chance and draw it contributes 15 per cent of the figure and the market 85 per cent; for goals and both-teams-to-score markets the figure is the market alone, because in our testing the model did not improve on the market there.
For each fixture the candidates are the full-match markets with a real bookmaker price of at least 1.20: a side to win, double chance, draw no bet, over and under 1.5, 2.5 and 3.5 goals, both teams to score, team goals, win-and-both-teams-to-score, win-and-over, and Asian handicap lines. A candidate the market rates below 40% is never published as our call.
The fixture gets the most likely candidate that passes the logic checks below, subject to daily limits on each bet type: no type may fill more than a fifth of the day's card, and double chance, over 1.5 and under 2.5 are held to a tenth. Without those limits the card would be mostly double chance and over 1.5 — high percentages at prices too short to tell the reader much. Over 2.5, match result and under 3.5 are preferred when they are close to the best option.
Draw and correct-score picks are not chosen as main predictions by this procedure; they have their own specialist pages and accumulators, because their honest probability is low by nature. Those pages list a draw from 27% and, for correct score, the most likely score that agrees with the fixture's main prediction.
A percentage alone is not enough; the pick also has to make sense for the match. A side to win needs the market to rate it at least 65%. Over 2.5 needs at least 55%, our model to expect at least 2.5 goals, and both teams to score to be rated at least 52%. Under 3.5 needs at least 65% and our model to expect fewer than three goals. Both teams to score needs at least 55% and the weaker side to be expected to score at least 0.9 goals. Double chance needs at least 75%, over 1.5 at least 75% and under 2.5 at least 60%.
Each of those rules was kept only after it held up in two separate periods of past results. When no candidate passes, the fixture keeps the pick from the earlier steps; if that pick itself fails a check, it is replaced by the best candidate that passes. A limit can stop a bet type from being used again, but it never pushes a fixture into a market that failed its checks.
Smaller competitions often have no bookmaker prices when picks are set. For those fixtures the main prediction comes from our team-strength model, using the same bet types and stricter minimum probabilities, because the model is not corrected by a market. No percentage is shown for them: the row reads no price. Where there is not enough match history for the model, the provider's advice is published instead.
Each fixture has one main prediction, but its other markets can appear on the market pages — over 2.5 on one page, both teams to score on another. Those entries are checked against each other and against the main prediction: everything shown for a fixture must be able to win together. A fixture never appears as over 2.5 on one page and under 2.5 on another.
On two-way market pages only the side the market rates as more likely is listed, at 50% or more. The two higher-risk pages are the exception by design: over 3.5 goals lists fixtures from 40% and under 1.5 goals from 30%, together with a supporting goal expectation, and they take precedence over their opposite pages for the same fixture.
The price attached to each pick is a real bookmaker's price, as offered when the pick went out. Freezing it at that moment is what makes a fair later comparison possible, because prices keep moving right up to kick-off.
Prices are also used to look for value. Where one bookmaker's price is clearly higher than the fair price implied by the rest of the market, the fixture can appear on the value bets page with the fair price and the size of the gap shown. Where a fixture has no price at all, no percentage is displayed — a probability with nothing behind it is worse than no probability.
The percentage beside a pick is the fair probability described above — not a claim about how often our picks win. It is banded for readability: high, medium and low, with a distinct state for fixtures that have no price. The band is a description of that single fixture and nothing more.
On the match insight page each pick is broken down into what drove it — recent form, attack and defence ratings, expected goals and the head-to-head record — with a short explanation that only uses comparisons in which the predicted side is actually ahead.
We do not publish hit rates, return-on-investment figures, profit claims or staking advice. Those numbers are trivial to present favourably and almost impossible for a reader to verify, and they are the single most common way prediction sites mislead people. Our picks and their published odds are on the site; anyone who wants to keep score can keep score.
We do not accept bets, process payments, hold funds or facilitate wagering of any kind. We are an analysis site. Nothing here is betting advice, financial advice or a guarantee of any outcome.
Predictions are generated once a night, before that day's fixtures, and the team-strength model is refitted shortly afterwards. Live scores and in-play events update continuously while matches are on. Standings, top scorer charts and team streaks refresh on their own slower schedules, since they change far less often.
Yesterday's predictions stay on the main page next to their final scores, and each covered fixture keeps a match page with the pick as it was published, the price and the final score.
If a fixture, price or score on the site is wrong because of a data error upstream, tell us through the contact page and we will correct the data. A published pick is not quietly swapped for a better-looking one: the pick stays as it was published.
18+. Betting involves financial risk. If gambling stops being entertainment, the responsible gambling page lists where to get help.