How the picks are made,
and how they are graded.

Every fixture is analysed from real match data, refreshed before each run:

• Recent form for both sides, weighted so newer matches count for more.
• Expected goals scored and conceded, split by home and away.
• Scoring and clean-sheet rates, goals per game, and league position.
• Head-to-head history between the two teams.
• Injuries and suspensions, and live team news when it is available.
• Market odds, used to check the model against the crowd, not to copy it.

Data comes from licensed sports-data providers and is cached briefly so a burst of users does not mean a burst of fetches. Nothing is typed in by hand.

Several independent AI models read the same data for the same fixture at the same time. Each returns its own view: the likely outcome in each market, a probability, and its reasoning. Their answers are merged by weighted consensus, so a pick has to survive more than one model to reach you. Where the models disagree, the disagreement lowers the confidence rather than being hidden.

For the daily board a further, separate pass reviews the merged candidates as a set. It can drop a pick or nudge a confidence figure within a narrow band. It cannot add a pick the models did not produce, and it cannot raise a figure above what the models supported.

The same models and the same process serve every plan. A free account is not analysed by a lesser system; it sees fewer picks.

The board publishes every morning. The headline board is the 5 picks that cleared the confidence floor across all leagues that day, ranked. Free accounts see the first 3, Pro sees all 5 with confidence, reasoning and key stats, and Elite adds an Extended Board of 5 further picks. Extended Board picks are tracked as their own segment and are not in the headline record.

On a thin day the board can be short. It is never padded to five with picks that did not clear the floor.

A confidence figure is the models' combined estimate of how likely the pick is to land, adjusted by the review pass. It is a probability, not a promise: a set of picks at 85% should land about 85 times in 100, and about 15 will not.

Whether that holds is checked, not assumed. The record page shows stated confidence against the actual hit rate in bands. If a band runs consistently under its stated figure, the figure is wrong and the process is adjusted. That comparison is the only honest meaning of the number.

• Each board is stored as published, with the time it was generated. It is not edited afterwards.
• After full time, each pick is graded automatically against the official result from the data provider. Match result, double chance, both teams to score, goals lines and correct score each have a fixed rule.
• Postponed, abandoned or unresolvable fixtures are voided. A void is not a hit and not a miss; it leaves the denominator.
• Hit rate is hits divided by settled picks. Pending picks are excluded until they settle.
• Every miss is listed on the record page next to the wins, with the final score.

The record is a complete account of the headline board since the season began, plus a frozen archive of the World Cup 2026 boards. It is not a claim about any single pick, and it is not a projection. A good month does not make the next one good.

Sample size matters. A hit rate over ten picks means little; over a few hundred it starts to. The page shows the counts next to every percentage so you can weigh them yourself.

You can track any prediction from an analysis. Where the fixture is covered by our data, the tracker settles it automatically from the final score and marks it verified. You can also settle a pick yourself; those are marked as self-reported and kept apart, because a record you grade yourself is an opinion.

The public Season Race ranks members by verified picks only, with a minimum number settled. Editing a settled pick removes it from the board. Appearing there is opt-in.

The Predictor League is a game, separate from the AI board: you call the result of each fixture in a gameweek before it locks. Points are awarded for the correct outcome, with a bonus for the exact score. Standings are recomputed after every settled fixture and prizes are granted automatically to the top three each gameweek.

LN AI is an analysis service. It does not take bets, hold funds or pay out. Nothing here is a guarantee of any outcome, and no confidence figure is high enough to stake money you cannot afford to lose.

The service is for adults aged 18 and over. If gambling is affecting you or someone close to you, help is available: in Nigeria, the National Lottery Regulatory Commission publishes responsible gaming resources, and international lines such as Gamblers Anonymous operate online. Set limits before you play, and stop when you reach them.

Questions about the method or a specific graded pick? Use the chat on any page, or reply to any email from us. Corrections to the record are made in the open and noted on the record page.