Methodology
How the numbers are made: the team-strength engine, the xP model, the price estimate, and how they perform against real results.
Every club carries an attack and a defence multiplier around the league’s average scoring rates (1.62 home / 1.32 away goals). The starting values are a joint Poisson maximum-likelihood fit over 1,864 Premier League and Championship results (2024–27), refreshed by the simulator’s weekly recalibration; on every load the same fit re-runs on this season’s finished fixtures, so the offline numbers fade as real results accumulate.
Scorelines come off a Dixon–Coles-corrected grid (the plain Poisson model misbehaves on 0–0 and 1–1). Win, draw, clean-sheet and 3+-goal probabilities are read straight off that grid. The same engine drives fixture difficulty, the Clean Sheet Matrix, Match Forecasts and the Season Simulator.
The fit is validated on held-out 2025/26 results: ranked probability score 0.2126 against 0.2133 for the unweighted fit. Small edges, honestly reported. The full method and code live in the open PL Simulator repository.
Since the season started, the Match Centre grades every pre-match forecast against the real result, including the misses, and the dashboard’s trust strip totals it up. When the model is wrong, you see it.
A player’s xP for a fixture blends the official expected-involvement rates (xG, xA per 90), minutes security (starts, availability flags), position-specific scoring including the defensive-contribution rules, and the team-level match forecast. A striker facing a leaky defence at home projects higher than the same striker away to the champions. Tap any xP number in the app for its workings.
FPL moves prices when net transfers cross a hidden ownership-scaled threshold. We approximate it, roughly 30% of current owners, floor 20k, and map progress through a logistic curve, capped 5–95%. It is an estimate and is always labelled as one; the real algorithm is not public.
Card-ban proximity uses the real cutoffs: 5 yellows by GW19, 10 by GW32, 15 across the season. Fitness comes from the official status flags and news feed. Deeper card analytics live in our companion app, Bookings Desk.
Match Centre with model verdicts · per-gameweek GW Debrief · suspension watch and push alert · personalised price alerts + evening risk warning · set-piece and start-likelihood chips · sortable transfer targets · shareable gameweek card · terminal redesign.
Is this the official Fantasy Premier League app?
No. Gameweek Edge is an independent companion that uses the official FPL data to give you deeper analysis. It isn't affiliated with, endorsed by, or connected to the Premier League or the official Fantasy Premier League game.
How is this different from the official app or a live-rank site?
The official app is where you set your team, but it has no predictions, no AI and a reputation for freezing on big gameweeks. Live-rank sites are great for your in-play rank and effective ownership, but they're ad-supported, desktop-first and stop at the numbers. Gameweek Edge gives you the live rank and EO, from the official data, ad-free and built for your phone. Then adds predicted points and an AI scout that tells you what to actually do next.
Do I need to pay?
No: the Free tier has everything you need to run your own team. Pro unlocks the AI scout, rival intelligence and the simulators for £3.99/month or £24.99 a season, and you can cancel anytime.
Which devices does it work on?
Any modern browser, on iPhone and Android. Add it to your home screen for a full-screen, app-like experience that works offline and can send you alerts.
Is my data safe?
Yes. We only use your public FPL Manager ID and the official data. Your account, watchlist and preferences are protected with row-level security, and we never sell your data or show ads.
How accurate is the AI?
The numbers, predicted points, optimal teams, probabilities, come from our own transparent models built on the official data. The AI explains the picks and is instructed to use only that data, so it won't invent stats.