FPL Expected Points

Minutes first. The model estimates how likely a player is to start, how likely he is to be subbed before 60 minutes, and how likely he is to miss out entirely. It reads the official availability flags and the chance-of-playing figures. Before the season and after a transfer, when there are no minutes to go on, it uses price and ownership as a market signal for expected role.

Per-90 rates, then the fixture. Goal, assist, save and defensive-action rates come from this season's xG, xA and event data, pooled with last season's numbers as a prior so early-season samples do not swing wildly. Each rate is adjusted for the opponent and venue. A Poisson team model sets expected goals for and against in every fixture, which gives the clean sheet probability and the expected goals-conceded penalty. Defensive contribution points use the probability of reaching the threshold: 10 clearances, blocks, interceptions and tackles for defenders, 12 including recoveries for midfielders and forwards.

Bonus and scoring. Bonus is estimated from each player's bonus history and BPS profile under the 2026/27 BPS rebalance. Every component is then passed through the official scoring table for the player's position, including appearance points, goals, assists, clean sheets, goals conceded, saves, defensive contributions and cards.

What to expect from the numbers. Even a perfect model has an error of about 2.8 points RMSE per player per gameweek, because football is random at the single-match level. Use these projections for direction: who to captain, which positions to strengthen, which price bracket gives the best return. Do not read 5.4 against 5.1 as a meaningful gap.