A team can outscore their opponents in three consecutive matches and still be playing worse football than their results suggest. Understanding the difference between xG and actual goals separates analytical bettors from casual ones.
Goals are the most visible football statistic, and the most misleading. Consider a striker who scores a 35-yard screamer into the top corner, a deflected cross that bounces off his knee and rolls over the line, and a tap-in from two yards out. All three are equal in the goals column. All three are wildly different in terms of actual skill.
At the team level, the same noise applies. A team can win 4-0 because a journeyman striker had the game of his career and hit the target four times with speculative shots. They can lose 0-2 while generating 2.4 xG against an opponent who scored twice from 0.3 total xG. The scoreline says the latter team is worse. The xG says exactly the opposite.
Over a large enough sample (typically 10+ matches, reliably around 20–30) actual goals tend to converge toward xG. The lucky and unlucky average out. The question is: how do you identify which teams are currently on the lucky or unlucky side of that distribution?
The difference between a team's actual goals and their expected goals is one of the most actionable signals in football analysis.
Goals > xG (Overperforming)
The team is scoring more goals than the quality of their chances warrants. Possible causes:
Signal: Regression risk. Results may deteriorate even if play stays the same.
xG > Goals (Underperforming)
The team is creating better chances than their goal tally shows. Possible causes:
Signal: Recovery candidate. Results should improve as variance normalises.
Football finishing quality (the ability to consistently score at rates above or below xG) is among the least stable individual statistics in the sport. At the team level, it is even less stable than at the individual level.
Research across Premier League seasons has consistently shown that teams finishing in the top half for xG-outperformance in the first half of the season significantly underperform their early-season goal rate in the second half, and vice versa. The conversion rate reverts.
Why finishing rates revert: the numbers
Low correlation; finishing rates are largely variance
Strong correlation; chance quality is a true skill signal
xG predicts future goals better than past goals in most studies
The practical implication: if a team is outperforming their xG by 0.5 goals per game over 8 matches (4 goals above expectation), that gap is almost certainly going to close. The team scoring 0.5 goals per game below their xG is almost certainly going to start getting luckier. The question is not if regression happens, but when.
The betting market is heavily influenced by recent results. A team on a 5-match winning streak will be priced as a strong favourite even if those wins were achieved on low xG, while their opponent played well but lost narrowly. The xG data tells a different story from the odds.
Fade overperforming favourites
Teams priced as strong favourites on the basis of recent results but whose xG significantly lags their goal tally are prime candidates to be faded. Their odds reflect their results; the market has not fully discounted the xG gap. Backing the underdog or the draw at inflated odds against a regression candidate exploits this inefficiency.
Back underperforming value plays
Teams whose xG significantly exceeds their actual goals (creating real quality but converting at below-expected rates) are often underpriced by the market. Backing them to improve in upcoming matches, particularly in good fixtures, is a consistent source of positive CLV before the market adjusts.
Check the xG differential, not just goals
The most complete picture is xG difference (xG scored minus xGA allowed). A team with +0.6 xG difference per game who has posted poor results due to finishing variance is fundamentally a strong team. A team with −0.3 xG difference who is winning through goalkeeping heroics and lucky finishes is fundamentally weaker than their record suggests.
Combine with xPts
Expected Points (xPts) calculated from the Poisson distribution using xG values shows how many points a team should have accumulated. Large gaps between xPts and actual points highlight the most extreme over- and underperformers; these are the highest-conviction regression plays.
For FPL managers, the xG vs goals gap is equally powerful, applied at the player level rather than the team level.
🔼 Buy: High xG, Low Goals
A midfielder or forward accumulating high xG per 90 but with fewer actual goals than expected is a prime buy. The goals are coming; they just have not converted yet. High xG + low ownership = differential with high ceiling.
🔽 Sell: High Goals, Low xG
A player whose goal tally significantly exceeds their xG is running hot. They are scoring more than chance quality warrants: long-range strikes going in, deflections, goalkeepers at fault. This rate will drop; sell before it does.
Combining xG with xA gives the fullest picture. A player with 0.35 xG + 0.20 xA per 90 over 8 gameweeks with only 1 goal return is significantly underperforming; they should have 3–4 goals already. This is exactly the kind of regression buy that wins at FPL.
xG is a powerful signal, but it has real limitations that should temper overconfidence:
xG measures the quality of chances created, which is a more stable and repeatable skill than the ability to convert those chances at a specific rate. Finishing quality fluctuates significantly from match to match; chance creation quality is more consistent. Over a large sample, actual goals converge toward xG, making xG a better predictor of future goal output than past goals.
As a rough guide, a difference of 0.3+ goals per game sustained over 5+ matches is significant enough to flag. Over 10 matches, a gap of 0.25+ per game represents a meaningful divergence from expected levels. Smaller samples have more noise: a single match difference of 1.5 goals over xG is meaningless without context.
At the elite level, some strikers show modest sustainable outperformance. Robert Lewandowski and Harry Kane are examples often cited in research. Their ability to pick the right shot and execute under pressure generates consistently above-expected conversion. However, the magnitude of sustainable outperformance is much smaller than casual analysis often implies, and most players who significantly exceed their xG in one season regress the next.
Yes. xGA (Expected Goals Against) vs actual goals conceded works the same way. A team conceding fewer goals than their xGA are getting lucky saves or fortunate misses; their defensive results should worsen. A team conceding more than their xGA is underperforming their defensive quality; they should concede less as variance normalises. Both are actionable signals for clean sheet markets and defensive player selection.
What is xG?
How expected goals are calculated: the foundation of this guide
What is xA?
Expected assists: the creative counterpart to xG
Poisson Distribution
How xG translates into scoreline probabilities and betting odds
Value Betting
How to turn xG overperformance identification into actual bets
npxG: Non-Penalty xG
Removing penalty noise for a purer measure of goalscoring quality
xPts: Expected Points
Turning xG data into expected league points: the ultimate table predictor
