Goal Involvements minus xGI
Goal involvements minus expected goal involvements is the over or under performance figure for a player's whole attacking game.
Why it matters. It is the one glance sustainability check. A large positive figure is riding good fortune, a large negative one is usually due an upturn.
One number for the whole attacking picture
Add goals and assists, add expected goals and expected assists, and subtract the second total from the first. What is left is the over or under performance across a player's entire attacking output, finishing and creating combined.
- Positive. Total goal involvements are running ahead of what the underlying chances, both taken and created, deserved.
- Negative. Total goal involvements are running behind those chances.
- Near zero. Output matches expected goal involvements closely, the normal state for most players across a full season.
The value of the combined figure is that finishing and creating can offset each other. A player scoring above their rate while creating below it can land close to zero here, even though neither half of their game is actually average. That is a feature, not a flaw: it answers one specific question, is total output ahead of or behind schedule, cleanly.
The convention is actual minus expected, same as the two deltas it combines. That is the opposite of the league table deltas, where xPts and xGD run expected minus actual. A positive number on a team's page there flags underperformance. A positive number here flags overperformance. Keep them separate.
Reading the combined figure in practice
A forward on +7 through the first two months. Goals and assists together are running well ahead of the chances involved. Verdict: this is a hot start, not a new level. Some of it will hold if the underlying expected goal involvements are also strong, but the gap itself should close.
A winger sitting on -6 by January. Finishing and creating are both lagging the chances they are generating. Verdict: check the split figures. If both goals minus xG and assists minus xA are negative, the whole attacking output is due an uplift, not just one half of it.
A player parked near zero across a full season, with strong expected goal involvements underneath. The most useful case on the page. Verdict: trust the expected total as the level of this player's attacking involvement, since the actual figures are tracking it closely.
Why this needs the biggest sample of all four
Every source of variance in the goals delta and the assists delta feeds into this one figure at once, so it settles more slowly than either half on its own.
- Under ten matches: almost meaningless. A single multi-goal game can swing the combined figure by several units.
- Half a season: usable as a rough guide, best paired with the split figures rather than read alone.
- A full season of regular starts: the point at which the combined figure is worth treating as a real sustainability check rather than a form snapshot.
Treat a large gap inside a small sample as a story still being written, not a verdict.
The misreading to watch for
"A figure close to zero means this player is having an average season." Not necessarily. It can mean two large, opposite gaps in finishing and creating have cancelled out. That is a different situation entirely from a player whose goals and assists have both tracked their expected numbers throughout. When the combined figure looks unremarkable, check the split deltas before concluding the season has been ordinary.
Common questions
- Why not read the goals and assists deltas separately?
- Because a player can be finishing hot and creating cold at the same time, and the two partly cancel out. The combined figure answers whether total output is ahead of or behind the chances involved.
- How big a gap is worth acting on?
- Judge it against minutes rather than in absolute terms. A gap of three or four across a full season is ordinary noise, while the same gap inside ten matches is a much stronger signal.
- Which way round is the subtraction?
- Actual minus expected, so positive means output is running ahead of chance quality. The standings deltas run the other way, expected minus actual, so keep the two conventions separate when reading across the site.
- Is a positive figure always good news for a player's rating?
- No, it is a warning to discount recent output slightly. It says results have outpaced the chances involved, which tends to correct. Rate the player on their expected goal involvements, and treat the delta as a note on how sustainable the recent numbers are.
- Does the combined figure hide anything the split deltas would show?
- Yes. A player can post a large positive goals delta and a matching negative assists delta, netting close to zero here while telling two very different stories underneath. Check the split figures if the combined one looks unremarkable but something still seems off.
- How much data does this need before it settles?
- More than either half on its own, because it inherits the sample size problem of both. Look for at least a full season of regular minutes before treating the figure as anything more than a snapshot of recent form.
Related terms
See xGI Δ across the Premier League
Every player, team and match measured on the chances created rather than the scoreline.
Last updated