FPL Analytics
The latest analysis on player trends and projections through the Premier League season - built for FPL managers.
What every FPL manager actually cares about
Ownership against output. Left is rare, up is productive, and the green corner is where a rank actually moves.
Every player, every metric, in one sortable table — then one click to his whole season, gameweek by gameweek, against what the model expected of him before each one.
Build a full fifteen under the real budget and club limits, for a specific gameweek, and see it break before the deadline does.
All 38 gameweeks with the blanks and doubles marked. Chip value is fixture structure, and the structure is knowable months out.
Who is walking into a kind run of fixtures and who is walking out of one, across three, five or eight gameweeks. Sort it either way.
Three days before every deadline: the highest-projected player in each position, the captain call, and last week’s misses first.
Projections updated 25 Oct · validated model
The Model’s Accuracy
- Rank correlation
- 0.36Against 0.27 for points-per-game and 0.25 for last-five form — all three measured on the same players, in the same gameweeks.
- Gameweeks it beat points-per-game
- 31/31Every gameweek tested, not the average of them. Against last-five form it wins 29 of 31.
- Calibration error
- 1.6%When it says 40%, it happens about 40% of the time — off by this much.
- Points of headroom left
- ≈0Per player per gameweek. Measured against the model's own distributions: what error remains is the sport being random, not the model being wrong.
Every one of these numbers is re-measured when the model changes, and the changes are written up as they happen — including the ones that failed. Read the blog →
This week’s write-up
What we believed, what we measured, what we did about it.
Part-time data analyst.
Full-time FPL manager.
I built this to answer my own questions properly, and kept building because the model got interesting. It is free, with no account, no paywall and no ads — and everything it gets wrong gets written up alongside everything it gets right.
More about the project →