Open benchmark
How accurate are FPL projections, really?
The short answer
Both teams are entered in the real game. Both are public.
Every figure below is read live from the official Fantasy Premier League entry API using the team id each side published themselves. Nothing is modelled and nothing is adjusted.
Neither entry id has been published for 2026/27 yet. Both rows fill in automatically — ours when the squad locks before the Gameweek 1 deadline, theirs when they post their reveal, which they have done in the fortnight before kickoff in each of the last three seasons.
Projections are frozen. Grading starts at the first whistle.
Nothing has been graded yet, so there is nothing here to claim. What exists is the commitment: the projections below were written down before the deadline and cannot be changed afterwards.
The first snapshot is taken automatically before the Gameweek 1 deadline.
What is in the log, and what is deliberately not
Four columns are graded on identical terms — the same players, the same gameweek, the same scoring:
- Onside v5 — the v5 engine number we actually serve on the site.
- Official FPL ep_next — the official game’s own projection, free and visible to every manager.
- Form (last 4 GW) — a naive baseline. If a model cannot beat this it is worth nothing.
- Season points per start — the other naive baseline.
Only players who reached 60 minutes are scored. That matters more than it sounds: include unused substitutes and any model can win by predicting zeros for the bench, which is how most published accuracy figures in this category are inflated.
What is not here, on purpose: Fantasy Football Hub’s per-player predicted points. That tool sits behind a paid subscription, and the only ways to obtain it are to breach the subscriber terms or to go through the unauthenticated cache that leaks their member content. We are not doing either. A page about checkable honesty, built on something taken, would be worth less than nothing.
Their results, by contrast, are entirely fair game and are in the panel at the top of this page — because they publish their AI team’s Fantasy Premier League entry id themselves and promote its finishing rank. That is a public number about a public team in a public game, and it is the comparison that matters more anyway. The projections column stays reserved, and the invitation below is open.
An open invitation to every FPL tool
Fantasy Football Hub, Fantasy Football Scout, FPL Review, anyone. If you publish projections, put them in the log:
- Send one CSV — player id, predicted points — before the gameweek deadline.
- We publish the file hash on receipt, so the submission is timestamped and unalterable.
- You are graded on exactly the population every other column is graded on: players who played 60 minutes or more.
- Your result is published whether it beats ours or not. So is ours.
- Withdraw whenever you like. The gameweeks you already submitted stay up.
Why this page names Fantasy Football Hub
They are the largest paid FPL platform in the UK and the natural point of comparison. Everything below is either their own published wording or a public record, and each is linked so you can check it rather than take our word for it.
- Their description of their model, in full: “The Algorithm is a statistical spreadsheet member’s tool developed by our resident statistician Carl Weeks. It uses various spread betting odds and expected bonus to determine the expected points for each player.” source
- We could not find a methodology page, a backtest, a calibration curve or a graded prediction log anywhere on their site. If one exists, tell us and we will link it here.
- Our equivalents are all public: the record, the calibration dashboard and the methodology. Onside v5 sits at MAE 0.86 across 51,518 out-of-sample predictions, against FPL’s own 0.896.
Snapshots are committed to a public git repository before each deadline, so the commit timestamp is independent proof the projection existed before the matches were played. Nothing on this page is recomputed after the fact.