Skip to content
FPLRogueResearch
Download on the App StoreGet it on Google Play
Settled result review

FPL GW3 projection accuracy scorecard

The accepted xpts_v1 release was frozen before the deadline, then compared with official points, minutes and starts after the Gameweek finished.

Snapshot2026-09-04-gw3
Likely players292
Likely-player MAE2.51 pts
xMins MAE24.1 min
No post-result reconstruction

Accuracy by cohort

CohortPlayersMAERMSEBiasWithin 2 ptsxMins MAEStart Brier
All captured players6521.492.21+0.6669.6%16.5 min0.106
Projected ≥45 minutes2922.513.04+0.9642.5%24.1 min0.185
Predicted top 25253.173.63+2.3724.0%18.2 min0.136
Highest pre-deadline xPts

Predicted top 10 outcomes

PlayerTeamPredictedActualErrorxMinsMinutes
#1 HaalandMCI10.109+1.1087.690
#2 B.FernandesMUN7.572+5.5788.090
#3 CherkiMCI7.113+4.1178.065
#4 GuéhiMCI6.998-1.0189.690
#5 O'ReillyMCI6.720+6.7281.30
#6 EnzoMCI6.553+3.5585.075
#7 SemenyoMCI6.246+0.2483.690
#8 FodenMCI6.151+5.1579.624
#9 SakaARS6.072+4.0769.289
#10 GvardiolMCI5.888-2.1270.890

How this scorecard is calculated

Point error is predicted next-Gameweek xPts minus settled official points. MAE averages absolute error; RMSE gives larger misses more weight; positive bias means overprediction. Minutes MAE compares xMins with played minutes. Start Brier score measures the probability forecast against the official start flag, where lower is better.