Pick your Fantasy Premier League team with a model, not a hunch.
Every fixture is priced by a Dixon-Coles match model fitted on three seasons of real results. Every player gets a points projection built from the actual scoring rules. Your squad is then chosen by an exact optimiser that solves the budget, position and three-per-club constraints together — not a greedy "best value" list.
Get going in three steps
Enter your fifteen
Go to My squad and search for your players, or press Auto-fill to start from the model's best team and edit from there. It saves, so you only do this once.
Importing by FPL team ID also works — but only after a gameweek deadline has passed. Before the season's first deadline the FPL API has no picks to return, so type them in.
Read your week
This week is the one screen that matters. It tells you which transfers to make and why, whether a −4 hit is worth taking, who starts, your bench order, who to captain, and whether to play a chip.
"Hold — make no transfer" is a real answer and it gives it often.
Act before the deadline
The countdown sits in the header. Make the moves in the real FPL site, then update My squad if you did something different, so next week's advice starts from the right team.
What each tab does
| Tab | Use it for |
|---|---|
| This week | Your weekly actions: transfers, the hit question, XI, bench order, captain, chips. |
| My squad | Enter or edit the fifteen you own, plus your bank and free transfers. |
| Inside information | On This week and My squad: tell the model what you know that it cannot — rested players, early returns, cameo appearances. |
| Players | Every player's projection, searchable and sortable. Click a name for the full points breakdown. |
| Fixtures | Win probabilities from the match model, and who blanks or plays twice. |
| Chips | When to play Bench Boost, Triple Captain, Free Hit and Wildcard. |
| Build from scratch | Design a squad under your own constraints — budget, must-haves, minutes floor, club spread. |
| Model | How well the match model actually forecasts, tested out of sample. |
Reading the numbers
Projections
Pts — expected FPL points. Over a horizon longer than one gameweek this is a decayed multi-week total, not a weekly score.
Vol ± — how much the actual score is likely to swing. It is the spread of the outcome, not the model's error. A premium forward is genuinely volatile.
Pts/£m — value. Useful for filling out a squad, misleading on its own: cheap bench players score well per pound and win you nothing.
Fixture colours
Difficulty uses FPL's own ramp, easiest to hardest: 1 2 3 4 5
An asterisk after an opponent means away. Hover any fixture for the win and clean-sheet probability behind it.
On Fixtures, the bar shows home win / draw / away win straight from the match model — not FPL's difficulty rating, which is shown separately for comparison.
Player flags
no-pl-data No Premier League record. The projection is inferred from similarly priced players — a guess, not a measurement.
new-club Summer signing. Their rates are real but were produced in a different team's system.
promoted-club Their club's rating is a league-average prior, not a measurement of that squad.
low-sample Too few minutes for their rates to mean much.
penalties First-choice penalty taker.
Chips
Payout — expected points gained, adjusted for how well that chip's forecast has historically predicted its outcome.
Chance it's the best week — simulated against every week left in the chip's window. With 19 weeks left, anything near 5% is no better than guessing.
Verdict — deploy when the payout beats what you would expect from the best of the weeks still available. That bar falls automatically as expiry nears.
How good is it?
What it does not know
Press-conference news — but you can tell it. The model cannot know what a manager said on Friday. Use Inside information, on both This week and My squad, to say a player is out, doubtful, easing back off the bench, or fit again. Your note overrides the data and flows straight into the transfer, XI, captain and chip advice — it will even recommend transferring in someone FPL still shows as injured, if you say they are back.
Promoted clubs. Their ratings are the pooled average of previously promoted sides. Championship form was tested as a predictor and rejected — it had no measurable relationship with subsequent Premier League level. Treat those players cautiously.
What you paid — unless you tell it. FPL's sell price is your purchase price plus half of any rise, rounded down. The public API will not reveal what you paid (that needs a logged-in session, and this tool never asks for your password), so enter it in the Bought for column on My squad and the sell price is then computed exactly. Leave it alone and it assumes you paid today's price.
Certainty. Single-gameweek football is mostly noise. The model has a real, measured edge over guessing, and it is a modest one.
Player, fixture and scoring data comes from the Fantasy Premier League game, published by the Premier League at www.premierleague.com, with all rights reserved by them. Match results from football-data.co.uk. This is an independent tool with no affiliation to the Premier League, Fantasy Premier League or any club — see .
Your fifteen
Add players
Must have — always included
Never pick — always excluded
Where the data comes from
Premier League / Fantasy Premier League
Player prices, positions, availability, fixtures and scoring data are sourced from the Fantasy Premier League game, published by the Premier League at www.premierleague.com. All rights in that material are owned by the Premier League and its suppliers, and are reserved.
These terms apply to you too. The Premier League's terms permit material from their website and app to be used for private and personal use, and to be passed on to others for their private and personal use, provided the source is credited. By using the projections and squad suggestions here you are asked to observe the same terms — read them at premierleague.com/terms-and-conditions.
This site is an independent tool. It is not affiliated with, endorsed by, or connected to the Premier League, Fantasy Premier League, or any football club. "Premier League" and "Fantasy Premier League" are the property of their owners.
Historical match results
Team ratings are fitted on match results from football-data.co.uk, which also supplies the bookmaker closing odds used to benchmark the model on the Model tab.
Historical per-gameweek player data used for backtesting comes from the vaastav/Fantasy-Premier-League archive.
Method
Match outcomes use a Dixon-Coles bivariate Poisson model, after Dixon & Coles (1997), Modelling Association Football Scores and Inefficiencies in the Football Betting Market. Squad selection is a mixed-integer program solved with HiGHS via SciPy.