A machine manages a real Fantasy Premier League squad. Every Friday it publishes its team and its predicted points — frozen before the deadline. Every Monday it grades every call in public, against the game's own experts. Even the wrong ones. Especially the wrong ones.
The XI, the captain, the transfers, and a predicted score for every player — published before the deadline and frozen. The experts' team goes on the board next to it.
Nothing to post. The machine doesn't get to explain itself mid-round, and neither do the experts. The predictions sit there, in public, aging.
Every call graded: CALLED IT or WRONG. The swing settles the week's round of Machine vs Experts. Wrong calls get the mark — no excuses, no deletions.
The flip is earned, never decorative. Upright, xP is the statistic the whole model runs on. When a graded call goes wrong, the statistic falls over and becomes a face with its tongue out — at us, not at you. When the machine is right, the mark stays a number.
An AI running a full model — market odds, form, fixtures, minutes — over a real £100m squad. It publishes what it expects before every deadline, to the decimal, and lets Monday say whether it was clever.
FPL's own Scout Selection — the official site's recommended best team, published weekly. The machine scores their XI with the same model, then the football scores both of us for real.
Owns the actual team. Clicks the actual buttons. Went 57–55 up on the experts in week one, then handed the keys to the machine — and has to live with whatever it does next.