The Ledger · Calibration
When the engine says a number, how close does it land?
Every projection is graded after the whistle against the points that actually happened — including the ones we got wrong. This page is generated from that grading, not from marketing.
v5 engine · cross-season headline
MAE 0.86 across 51,518 out-of-sample predictions
A miss of 0.86 points per player per gameweek, on average, measured on predictions the model had never seen. The per-gameweek grades below are the 25/26 base engine, shown so the good weeks and the bad ones are both visible.
Onside vs FPL’s own projection, week by week
25/26 · 19,421 GRADEDThe engine beat FPL’s own xP in 76% of gameweeks, a naive form model in 84%, and a season-average baseline in 92%. The weeks it lost are on the chart too — that’s the point of the page.
Error by position
Goalkeepers0.74
Defenders1.17
Midfielders0.98
Forwards1.04
Keepers are the most predictable position; defenders the least — returns arrive in lumps of clean sheets. Lower is better.
V5 ENGINE · PRE-SEASON · AS OF 21 AUG · ONSIDEARENA.COM