The data

• Fantasy Premier League’s free public data: players, prices, ownership, fixtures, results, minutes, and expected goals and assists.
• Past seasons: every player’s gameweek scores going back to 2016/17, used to train the expected points model.
• Nothing else. No betting odds and no paid stats feeds.

The goals model

For every match it estimates how many goals each side will score, then turns that into win, draw and loss chances and a chance for every scoreline.

• Starting point: Fantasy’s own attack and defence ratings for each club, home and away.
• This season: each club’s goals and expected goals, for and against. Expected goals count for more than actual goals, because they’re steadier.
• Blending: early on the model leans on Fantasy’s ratings. As the season goes on, this season’s form counts for more; after eight matches it counts as much as the ratings.
• Home advantage: home sides are expected to score more.
• Scorelines: goals for each side follow a Poisson spread, with a small correction so low-scoring draws like 0–0 and 1–1 come out a little more often, as they do in real football.

How the games use it

• Last Fan Standing: win chance for each club. Safe is 60% or more, Fair 45–60%, Risky under 45%. The planner looks six gameweeks ahead and spreads your unused clubs across the weeks to give the best chance of surviving them all.
• Six: the six matches are the closest of the gameweek, the ones with the smallest gap between home and away win chance. The likely score is the single most probable scoreline. The safest double-up is the match where the model is surest of the result.
• Scorers: a player’s share of his club’s expected goals this season, times how many goals the club is expected to score this week, turned into a chance to score at least once.
• Captain Call: the same idea using goals and assists, to rank the biggest threats.
• Season picks: current points plus the expected points from every match left.

Expected points

Exp. pts is a forecast of a player’s FPL points in the next gameweek. It comes from a machine-learning model trained on past seasons’ gameweek scores, learning how points follow from what came before.

• What it looks at: recent points, minutes, expected goal involvements, goals, assists, clean sheets and bonus, plus price and whether the next match is at home.
• How often: it retrains once a week when a new gameweek’s scores are in.
• One week only: it forecasts the next gameweek, not the next five.

The Pitchline score

A simpler, hand-built score used where a longer view helps. It adds up a player’s form, expected goal involvements, ICT index and points per million, rewards an easy run over the next five fixtures, and takes points off when he’s flagged as doubtful or out. It also leans a little towards popular or rarely owned players, depending on how you play.

Rankings and captain picks sort by expected points when it’s available, and use the Pitchline score otherwise. Metrics explains every column.

Fixture difficulty

Difficulty is Fantasy’s own 1 (easy) to 5 (hard) rating for each match, not a Pitchline model. Outlook is the same rating turned round, so a higher number is better.

The Heroes

Clark, Diana and Peter make their game picks before each deadline from the same goals model, each in their own style: Clark goes for the safest option, Diana balances risk, and Peter looks for picks few others make.

Their FPL squads work differently. Transfers are made before the deadline, but each Hero’s starting XI, captain and chips are chosen after the gameweek, from the points players actually scored. So their FPL totals show what each style’s squad could score, not a forecast you could have copied.

Checking the results

Before each gameweek, Pitchline saves every player’s expected points. Afterwards, This week compares them with what happened: the closest calls, the near misses and the big surprises. The saved forecast is taken a few days before the deadline, so it doesn’t include late team news.

What it can’t see

• Team news: expected points don’t yet drop for an injury flag. Check flags yourself (see Injury flags & team news).
• Fixtures in expected points: the forecast doesn’t use fixture difficulty, and doesn’t double for a double gameweek. Check the fixtures too.
• Transfers between clubs: a club’s expected goals count only the players there now.
• The unexpected: red cards, injuries in the warm-up, and plain luck. Every number here is a hint, not a promise.