Prediction ledger · 2026/27 season

Locked before kickoff. Graded after.

In the final hours before every gameweek deadline, the projections on predicted points are frozen — saved to a file that is never edited again. Once the gameweek finishes and the official points are settled, the frozen numbers are graded against what actually happened, alongside two simple yardsticks anyone could have used instead: each player's season average and his recent form, also frozen at the deadline. No cherry-picking, no after-the-fact edits — bad weeks land on this page exactly like good ones.

The ledger starts at the GW1 deadline (21 August). The first freeze happens in the final two hours before kickoff, and the first grades land here once the gameweek's points are official — usually the following morning. Until then, the method's proof is the full walk-forward test on last season, summarised on the predicted points page: the engine was more accurate than season-average and recent-form yardsticks across 31 straight gameweeks.

How the grading works

“Predicted starters” are players projected for 60+ minutes at the freeze. Their predicted scores are compared with the official points once bonus is settled. The two yardsticks — season average and recent form — use FPL's own published figures, frozen at the same moment, so all three strategies face the same week with the same information. One football match is mostly luck; the question this page answers is whether the engine is closer than the obvious alternatives, week after week.

FPL is a free fantasy game — this is squad research for your fantasy team.