The model's record
Every gameweek, inside the day and a half before the deadline, this site writes down the model's Team of the Week, its captain pick and its five most likely price moves. After the gameweek is scored they are graded against what an average manager scored, against an XI picked on form alone, and against every player who took the field. Everything below is that record, including the weeks it went badly.
Loading the record…
The season so far
The most recent gameweek
Gameweek by gameweek
How this is kept honest
Picks are recorded before the deadline, or not at all. A scheduled job writes the gameweek's file inside the 36 hours before the deadline and stores both times, so the gap is on the page. It refuses to write after a deadline and refuses to overwrite a gameweek that exists. There is no flag that turns either refusal off, and the file is committed to a public repository, so its history is timestamped evidence. A missing gameweek means the job did not run or the feed was down; it is never filled in later.
The picks are the ones the app showed. The job runs the same functions the app runs, lifted out of the same file, not a copy of them.
The Team of the Week is graded on the eleven's official points, undoubled. FPL's average manager score includes captains, chips and automatic substitutions, so the two are not the same measure. The XI with the model captain doubled is shown beside it; the honest headline is the plain sum. The form XI is the best legal eleven by FPL's own form figure, chosen at the same moment as our picks, never afterwards.
The captain is graded by rank. Where he finished by official points among every player who played that gameweek. Rank one means the model named the top scorer.
A price call is a hit when the price moved the way the model said between the record and the grade. A price that did not move is a miss, not a pending call.
"Not graded" means the feed did not publish the input. Nothing on this page is estimated to fill a gap.
None of this makes a projection a promise. A model that beats these bars over a season is one worth reading alongside what you know; it does not know what happens on Saturday.