01
Plan this Gameweek and look ahead two more before the deadline.
Loading Gameweeks…
02
Computing
FPL player predictions
Expected points (xPts) from form, minutes, FDR and live official FPL data01
—
Projected Gameweek points
— xPts
—
Captain (armband)
15 players
—
Squad value (£m)
<10% TSB
—
Differential picks
Past + live
Form & minutesFixtures & FDRPrice &
ownershipxPts model
Ridge · XGBoost · CatBoost · MLP ensemble
03
OFFICIAL / FIXTURES
Premier League fixtures & FDR
Kick-off, home or away, and official Fixture Difficulty Rating at a glanceLoading Premier League fixtures…
04
AI squad selection
15 picks, captain and vice-captain for the selected Gameweek
2 • 5 • 5 • 3
Projecting the Gameweek…
| Rank / player | Pos | Opponent | Price | TSB % | Availability | xPts |
|---|
05
How the FPL model works
Data, validation, measured accuracy, and known limitsThe technical report
OpenFPL was measured on a full Premier League season held out of training — 29,747 player-gameweeks. The report documents the four-model ensemble, how it was validated, exactly how accurate its expected points proved to be, and where it is known to be weak.
- Holdout RMSE
- 1.92 points per player-gameweek
- Rank correlation
- 0.70 mean Spearman ρ, all 38 GWs
- Squad points
- 70.6 per GW vs 56.0 for FPL's xP
- Gameweeks won
- 31/38 against a 5-GW form squad
Fork it, test it, make it better
OpenFPL is open source. The models, the training pipeline, the backtests behind every figure in the report, and this app are all in the repository — fork it to run your own experiments, open an issue if a number looks wrong, or send a pull request. Contributions of any size are welcome.
06
FPL questions, answered
xPts, captaincy, differentials, flags and how the team rating is scoredHow does OpenFPL predict FPL points?
An ensemble of four models — Ridge, XGBoost, CatBoost and an MLP — reads each player's recent form, minutes, availability, ownership, price and fixture context from the official FPL API, then projects expected points for the upcoming Gameweek. The squad selector applies the official rules: two goalkeepers, five defenders, five midfielders, three forwards, and a maximum of three players per Premier League club.
What does xPts mean in FPL?
xPts is expected points — the average FPL score a player is projected to return in a Gameweek, before the captain's double. It blends the chance of starting, appearance points, attacking returns, clean sheets and bonus points into one number you can rank players by.
Who should I captain this Gameweek?
The armband goes to the highest projected starter in the scout squad, shown on the Captain card with the vice-captain marked VC on the pitch. Captain points are doubled, so the pick is the single biggest swing in your Gameweek — always confirm the player is fit and starting before the deadline.
What is a differential in FPL?
A differential is a player owned by less than 10% of managers (TSB). Differentials gain rank when they return and cost you little when they blank, because most of the field does not own them. OpenFPL counts the differentials inside the recommended squad each Gameweek.
Does OpenFPL account for injuries and flagged players?
Yes. Availability comes straight from official FPL, so injured, suspended and doubtful players carry their flag and are filtered out of selection. The model can only rank what has already been published, so check late team news and press conferences before the deadline.
Does OpenFPL consider my budget, transfers and chips?
No. Squad selection is deliberately budget-free: prices and team value are shown for context but do not affect projections or picks. Transfers, hits, and chips such as Wildcard, Bench Boost, Triple Captain and Free Hit are your call — use the projections to inform them.
How is my FPL team rated out of 100?
Enter your public FPL team ID and OpenFPL reads your latest published picks, projects them for the selected Gameweek, and scores them against the AI benchmark squad. The rating combines your starting XI, your captaincy call and the availability of your players.