FPL Derby
01:30
🤍 💙 FPL 26/27 🤍 💙 · Gameweek 9

Accuracy

Every scoring pass is recorded before a gameweek and graded against what happened after it. The benchmark is FPL's own expected points — a weighted composite that cannot beat one published number is not earning its complexity.
Runs graded
5
5 recorded
Beats naive
1/5
vs FPL's own expected points
Best lift
5.97
top band minus bottom, points

Graded runsrank correlation with actual points

GameweekPlayersCompositeNaiveLift
GW8
2026-10-30 12:00
3060.5350.5914.52loses to naive
GW7
2026-10-23 12:00
2850.6480.7335.05loses to naive
GW7
2026-10-22 12:00
2860.6490.7335.07loses to naive
GW6
2026-10-16 12:00
2790.7450.7375.52beats naive
GW5
2026-10-09 12:00
2630.6830.7555.97loses to naive

Score bands · GW8average points actually scored, best band first

BandActual pointsAvg
Band 1
61 players · avg score 62.1
5.44
Band 2
61 players · avg score 46.1
3.98
Band 3
61 players · avg score 38.2
2.41
Band 4
61 players · avg score 30.9
1.82
Band 5
62 players · avg score 23.0
0.92

How this is measuredwalk-forward, no look-ahead

A scoring pass is recorded before a gameweek, with the weights and coverage that were live at the time — a score is only interpretable next to the formula that produced it, and that formula changes as coverage improves. Once FPL marks the gameweek checked, what every player actually scored is attached to that run and never revised. Composite is the rank correlation between the model's score and actual points; naive is the same for FPL's own expected points alone. Lift is the average points of the top score band minus the bottom. Rows are written only by a scheduled call with a shared secret; this page reads them.
Not affiliated with the Premier League. Player and league data from the official Fantasy Premier League API.