Expected goals (xG), the metric that assigns each shot a probability of scoring based on historical shots from similar situations, was designed to isolate chance creation from chance conversion — and that is exactly why it hides finishing quality. xG prices the shot, not the strike: a header from six yards with a defender on the shooter's back and a keeper narrowing carries the same value whether the finish is dinked over or blazed wide. Model evaluations published through the 2010s and early 2020s, including Michael Caley's public xG work and Opta's model documentation, converged on a figure analysts still cite: the best explainable factors — location, body part, assist type, defensive pressure — leave a large share of shot-to-shot variance unexplained, which is precisely where finishing skill lives.
What does xG actually measure — and what does it ignore?
A standard xG model ingests shot location, body part, pattern of play, assist type and, in more advanced versions, goalkeeper and defender positions at the moment of the strike. It outputs a number between 0 and 1: the fraction of comparable historical shots that became goals. Summed over a match or a season, xG is a strong estimator of the quality of chances a team or player generated.
But the model's inputs are deliberately limited. It does not know where in the goalmouth the shot was aimed, how quickly the ball arrived, whether the shooter opened his body to open the far corner, or whether he shot across the goalkeeper's momentum. Two shots with identical xG can be a guided finish into the bottom corner and a panicked toe-poke. The metric's strength — averaging away detail — is structurally the same as its blindness.
Why do finishing overperformance and underperformance mislead?
The most quoted xG statistic in fan discourse is a striker's goals minus xG: finish above your xG and you are clinical, below it and you are wasteful. Over a single season, this comparison is mostly noise. Statistical studies of shot data have repeatedly shown that season-level goal-to-xG differences regress heavily toward the mean, and that a meaningful finishing signal needs multiple seasons and hundreds of shots to separate skill from variance.
The misreading runs both directions. A forward who scores 24 goals from 18 xG is often called a poor finisher waiting to regress — sometimes correctly. But a small group of elite finishers sustain positive differentials across many seasons, which is the analytical fingerprint of skill. Conversely, players who convert cleverly — shooting earlier, aiming far post, forcing the keeper to set his feet — can underperform raw xG while producing better shot selections than the model credits. The differential is a question, never a verdict.
What do Post-Shot xG and shot placement reveal?
The bridge between xG and true finishing analysis is Post-Shot xG (PSxG), sometimes called xG on target. PSxG re-prices the shot after contact, using where the ball went: placement, height and pace relative to the goalkeeper's position. A shot placed inside the post at hip height carries a high PSxG regardless of its pre-shot xG; a tame central effort carries a low one.
| Metric | Measured at | What it isolates |
|---|---|---|
| xG | Before the shot | Chance creation quality |
| PSxG | After contact | Placement and execution |
| Goals − PSxG | Goal event | Goalkeeping, plus finishing variance |
| Goals − xG | Goal event | Mixed creation and conversion signal |
For goalkeepers the logic inverts: PSxG minus goals conceded measures shot-stopping. For finishers, the gap between a shot's xG and its PSxG — was the attempt directed well once taken? — is the closest widely available proxy for technique. Even PSxG, though, ignores shot selection: a striker who declines low-value shots to manufacture better ones shows up as a modest creator rather than the intelligent processor he is.
How should analysts judge a finisher properly?
A defensible finishing assessment layers at least four signals. First, shot volume and location profile: does the player generate attempts from high-xG zones himself, through movement and first-touch setup? Second, multi-year goals-minus-xG trends: sustained differentials over several hundred shots carry real signal. Third, PSxG-to-xG ratios for placement quality. Fourth, qualitative shot types — one-touch finishes, far-post aiming, first-time strikes from cut-backs — which scouts cross-check against the numbers.
This layered approach matters practically. Transfer valuations that lean on a single hot season of xG overperformance routinely overpay for variance, while undervaluing finishers whose value shows up in shot selection rather than conversion. The metric did not fail in those cases; it was simply asked a question it never claims to answer.
Where does xG still earn its place?
None of this makes xG a broken tool. Over a season, team-level xG difference predicts future results better than goals or shots, which is why clubs use it to strip luck from form tables. It excels at evaluating chance creation, attacking patterns and defensive generosity — the supply side of scoring. The honest framing is division of labour: xG describes the kitchen, PSxG describes the plating, and only long-run, multi-season evidence can say anything about the chef.
How should readers use xG responsibly?
The responsible use of xG treats it as a baseline, not a verdict. Single-match numbers are noisy and say little, while season-long gaps between goals and xG are more informative, and even then they should prompt questions rather than conclusions: did the finisher change, did the chance profile change, did the team's shot selection improve. When a striker overperforms for two consecutive seasons, finishing skill is a reasonable hypothesis; when the gap opens suddenly, variance and role changes are likelier explanations. Comparing players on xG per shot rather than total xG also isolates chance selection from volume, which is often the real story. The metric remains one of the best tools supporters have for seeing past scorelines, precisely because it knows what it does not know. Used as a measuring stick for chance quality and a prompt for better questions, xG sharpens analysis; used as a full accounting of finishing, it quietly erases the very skill it appears to measure.
For more context, read How PPDA measures pressing intensity and what it misses.
For more context, read set-piece coach.
For more context, read How goalkeeper distribution quietly rewired build-up play.