xG (expected goals) is one of the most quoted numbers in modern football. It can genuinely help you understand chance quality and repeatable performance, but it also creates plenty of bad takes: “we deserved to win”, “they were lucky”, “he’s finished”, all based on one match’s xG. The truth sits in the detail: how xG is built, what it ignores, and how quickly it becomes noisy when you treat it like a final verdict.
What xG actually measures (and what it does not)
At its simplest, xG is a probability assigned to a shot: a number between 0 and 1 that estimates how often a similar chance becomes a goal. Providers train these models on large historical shot databases and use information about the attempt—things like location, angle, body part, and the build-up context—to estimate how “scorable” the shot was. When you add up every shot’s xG in a match, you get a rough expectation of goals from the chances created, not a promise of what “should” have happened.
This point matters because many people talk about xG as if it measures overall performance in every phase of play. It doesn’t. Traditional xG starts when a shot is taken, so it cannot fully capture how a team played without the ball, how well they progressed through midfield, or how often they reached dangerous areas but failed to shoot. A side can control territory, force errors, and still rack up modest xG if they settle for blocked attempts or take shots from poor angles.
It’s also worth knowing that xG is not one universal number. Different providers use different inputs and different training data, so two respected sources can assign different values to the same match. Even a small modelling choice—how to treat defender pressure, whether to include goalkeeper position, how to label “assists” or “crosses”—shifts the output. That’s why it’s sensible to compare like with like (same provider, same competition, same definitions) before drawing strong conclusions.
Pre-shot xG vs post-shot xG: the detail that changes the story
Most match graphics show a pre-shot xG: the chance quality at the moment of the shot, based on the situation that led to it. This is the version that helps you separate “we created good looks” from “we had lots of hopeful efforts”. Two teams can each have 1.5 xG and still look very different—one might have five medium chances, the other one huge chance plus a pile of low-probability shots.
Post-shot xG (often called PSxG) adds information about where the shot actually went—its placement towards the corners or straight at the keeper. That makes it useful for evaluating finishing and goalkeeping: the same shot location can be hit poorly or perfectly, and post-shot xG tries to reflect that. If a goalkeeper concedes from a shot with very low post-shot xG, it hints at a mistake; if a striker consistently beats the keeper with high post-shot difficulty, it suggests elite shot placement rather than merely being on the end of good chances.
Be careful, though: post-shot xG can flatter or punish players in small samples. A couple of rockets into the top corner will spike a forward’s PSxG, and a couple of deflections can make a keeper’s numbers look ugly. The right way to use it is over longer runs and with video checks—especially when rebounds, screens, and slight touches are involved.
How to read xG properly: a practical match and season checklist
Start with shot-by-shot context, not the headline total. One 0.75 chance created by a cut-back to the six-yard box is not the same as fifteen 0.05 efforts from outside the area, even if the totals match. When you look at a match, ask: how many “clear” chances were there, and where did they come from—quick transitions, sustained pressure, set pieces, or individual errors?
Split the data into meaningful buckets. The most basic split is open play vs set pieces. Set-piece xG is real, but it behaves differently: it depends heavily on delivery quality, rehearsed routines, and referee decisions (especially penalties). When comparing teams, it can be fairer to look at open-play xG separately, then add set pieces back in once you’ve understood the style and the repeatability.
Watch out for penalties and their outsized effect on totals. Many models treat every penalty as a fixed value around three-quarters of a goal. One penalty can swing a match’s “xG story” more than twenty minutes of open play. In analysis, people often look at non-penalty xG (npxG) to avoid letting a single decision dominate the interpretation.
What to do when xG and the scoreboard disagree
First, check if the xG gap is driven by a small number of events. If the losing team had one massive chance (say 0.8 xG) and missed, the match will look “unfair” in xG terms, but football is allowed to be decided by a handful of moments. That’s not luck in the mystical sense; it’s just how low-scoring sports behave.
Second, consider finishing and goalkeeping variance. A side can overperform xG for a few matches because a striker is in a hot streak or the keeper is making match-defining saves. Over a full season, teams tend to drift closer to their underlying chance quality, but individual players can maintain finishing edges—especially if they consistently take shots from favourable zones or have outstanding shot placement (which post-shot metrics capture better).
Third, ask whether the match plan traded shot quality for control. Some teams accept low-probability shots as a way to keep pressure and prevent counters; others avoid shooting unless the chance is clean, which can make their xG per shot look better while their total shots look lower. xG doesn’t judge tactics by itself—it only describes the shots that happened.
Why xG often misleads: common traps even smart fans fall into
The biggest trap is treating xG like a “deserved goals” scoreboard. xG is an estimate built from historical averages, not a law of nature. If a match ends 1–0 with the loser on 2.1 xG, it doesn’t automatically mean the loser played better. It might mean they created chances and failed to take them, or it might mean their xG came from low-value shots plus one big miss, while the winner created fewer but more controllable moments and defended the box well.
Another trap is ignoring provider differences and data quality. The model can only be as good as the event data and labelling. If defensive pressure or goalkeeper position is not captured (or captured inconsistently), two similar shots might be treated as identical when they were not. Some modern updates explicitly add more contextual factors to reduce these blind spots, which is part of why xG values can change when providers revise their models.
A third trap is overreacting to tiny samples. One match of xG tells you very little. Five matches tell you a bit more. Over 20–30 matches, you start to see the stable shape of a team: whether they consistently create good looks, whether they allow dangerous shots, and whether their style produces repeatable chances. Even then, you should still cross-check with video and opponent strength.
Better questions to ask with xG (so you don’t get led astray)
Instead of “Who deserved to win?”, ask “What kinds of chances were created, and are they repeatable?” A team living off chaotic rebounds and last-ditch blocks may generate messy xG that swings wildly week to week. A team generating regular cut-backs and one-v-ones is usually building something more sustainable.
Instead of “Is this striker bad because he underperformed xG?”, ask “What shots is he taking, and from where?” A forward forced into low-quality shots will underperform expectations of casual viewers even if his finishing is fine. Also check whether he takes penalties; separating non-penalty numbers is often a fairer view of open-play contribution.
And instead of “xG proves the narrative”, treat it as a tool that needs context. The most useful habit is a simple routine: check penalties, check open play vs set pieces, look at the top three chances in the match, and compare the picture to what you actually saw. Used that way, xG helps you understand football rather than argue about it.
