Everything on this page is derived from our own player data: a team's expected goals in a gameweek is the sum of its players' expected goals (xG) from the official FPL feed. It is not a bookmaker's or a data provider's team xG, levels can differ from what other sites show, even when the ordering agrees.

xGF, xGA and the double-gameweek rule

xGF (expected goals for) is exact: player xG is a gameweek total, so summing it always gives the team's gameweek total. xGA (against) needs to know which match the goals were expected in, and when a team plays twice in one gameweek, the per-match split simply isn't in the data. Those fixtures are excluded from xGA and xPts and counted in the coverage note, rather than being approximated. The same limitation hides the second match's real score in a double gameweek, the affected rows carry a ^*.

Expected points (xPts)

For each cleanly attributable fixture we treat both sides' xG as goal rates and compute the chance of every scoreline (a Poisson model). That gives a win/draw/loss probability per match, hence expected points: 3 × P(win) + P(draw). A team's xPts is the sum over its attributable fixtures; x# is the table ranked by xPts.

Reading the Δ columns

Both Δ columns compare reality with the underlying numbers, and both use the same colours:

• Green, the xG numbers rate this team better than its results have paid it so far. Underrated; things tend to improve.
• Red, the team has banked more points than its xG earns. Overperformance like this usually regresses.

The forward model (Goals / CS% on the Fixtures page)

Upcoming-fixture expected goals come from each team's derived attack and defence strength (average per-match xG for and against, relative to the league) plus a home/away factor measured from the data. Early in a season, when there are few matches to measure, last season's strengths are blended in and phased out over the first eight matches; newly promoted clubs start at the league average until they have played enough. Nothing is ever invented, when there is no sample at all, the model says so instead of showing numbers.

What this is not

This is a model, not a prediction. xG measures chance quality, not finishing skill, goalkeeping, game state or luck, a team can legitimately beat its xG for a long time. Treat the table as one lens on the season, not a verdict.