Expected Goals (xG) on the Scoreboard: What the Number Really Tells You
Why a 2-0 Win Can Hide a Shaky Performance
Image: DSC2582, by hasgeek, CC BY 2.0
Picture a midweek league game that finishes 2-0. The winning side scored twice, kept a clean sheet, and the table shows three points. Now look at the shot count: 6 attempts to 19. The losing team hit the post, forced two saves from close range, and spent long spells camped in the final third. The winners scored from a deflected cross and a penalty.
The scoreboard records outcomes. It tells you what went in, not how often a team got into positions where goals usually follow. Two finishes from twelve yards and eighteen attempts from thirty yards can produce the same scoreline, and the second performance is far more repeatable than the first. Over 38 games, chance quality tends to show up in the points column. Over one match, it often doesn't.
That gap between what happened and what was likely to happen is where the next number on the scoreboard comes in.
This article sets out to make Expected Goals readable at a glance. First comes a plain-language definition, short enough to explain to someone in a pub without losing them. Then the mechanics: which inputs feed a shot's value, why distance and angle matter more than the highlight reel, and where the models differ. From there, the practical part. A high Expected Goals figure with no goals on the board points in one direction; a low figure with three goals points in another, and neither is a verdict on its own. You will see why one match rarely proves anything and what a run of games can show. By the end, the number beside the scoreline should read as context rather than a puzzle.
Measuring Chance Quality, Shot by Shot
Image: DSC2816, by hasgeek, CC BY 2.0
Strip it back and Expected Goals is a probability attached to a single shot: the chance that this attempt, from this position, with this body part, becomes a goal. The number comes from history. Analysts feed thousands of comparable shots into a model and count how often they were converted. A penalty lands around 0.76, meaning penalties like it go in roughly three times out of four. A header from six yards sits near 0.4. A hopeful effort from 30 yards is about 0.03, which is to say it almost never works.
Every value is per shot, never per match. Team Expected Goals is simply the sum of those individual probabilities, so ten half-chances can add up to more than two good ones. The figure describes the situation, not the people in it. It does not know that a striker is out of form, that a defender slipped, or that the goalkeeper is having the game of his life.
The model starts with geometry. Distance to the goal and the angle subtended by the posts do most of the early work, because a shot from twelve yards out in the middle of the box is obviously a better bet than one from a tight angle near the byline. Then come the modifiers: foot or head, the kind of pass that created the chance (a through ball or cutback beats a hopeful cross), how many defenders were between shooter and goal, and whether the move came from open play or a set piece.
Each of those inputs shifts the estimate up or down, and the weights attached to them are learned from historical shot data rather than guessed. Providers keep retraining as more matches are logged, so the same shot can drift in value from one season to the next.
That last point matters when you compare numbers. Opta, StatsBomb and the rest build their own models with their own inputs and their own training sets, so a team's 1.8 on one site may read 1.6 on another. Check whose model you are reading before drawing conclusions.
- Distance and angle to goal: the closer the shot, and the wider the view of the posts, the higher the value.
- Body part: feet generally rate above headers. The further out the attempt comes from, the wider that gap.
- Assist type: through balls and cutbacks lift the number. Hopeful crosses and long punts drag it down.
- Defensive pressure: count the defenders between shooter and goal, then ask whether the shooter was closed down at the moment of contact.
- Set piece or open play: corners, free kicks and penalties carry their own baselines, separate from the run of normal play.
Reading High and Low xG Without Jumping to Conclusions
A team that racks up 2.4 Team Expected Goals and scores once has not been outplayed. It has been let down at the final step. The chance creation part of the job was done: 2.4 is a healthy afternoon's work, the sort of number that usually produces two or three goals. When the scoreline reads 1-0 the other way, the gap between the two figures is the story, and most of the time it closes on its own. Conversion rates over a handful of games swing wildly and then settle.
That said, "bad luck" is not the whole explanation, and treating it as such gets managers sacked. Some forwards are genuinely below average finishers, and the pattern follows them from club to club. Goalkeepers have hot seasons. A side that keeps posting high xG with low goals may have a striker problem, a shot-selection problem, or both.
The fix is sample size. One match tells you almost nothing. Five to ten matches of the same pattern tells you something worth acting on.
The mirror image is the one that should worry you more. A side that puts up 0.9 Team Expected Goals and wins 3-0 has scored three times from roughly one goal's worth of chances. That is not a blueprint; it is a team living on borrowed finishing. Of the two mismatches, this is the one that more often comes before a run of poor results, because there is no underlying process to fall back on when the goals dry up.
The caveat matters, though. Elite finishers and set-piece specialists can beat their xG for long stretches, and some teams are simply built to score from the situations the model rates modestly.
So check where the goals came from. Penalties, deflections and long-range strikes inflate a scoreline without telling you anything repeatable. Strip those out and ask what was left: if the answer is very little, the next month will be harder than the last one.
Three Rules for Reading xG Without Overreacting
- Judge over five to ten matches, not one, because a single game's xG swing is mostly noise rather than a trend.
- Check whether penalties, deflections or long-range strikes are padding the scoreline, since those inflate goals without a repeatable process behind them.
- Compare a team to its own xG baseline rather than a league average, because styles differ and a low-xG side can still be performing to plan.
Where Expected Goals Stops Being Useful
The shot model has no idea who is in goal. A 0.30 chance struck past a fit first-choice keeper and the same 0.30 chance struck past an emergency stand-in are scored identically, because Expected Goals treats finishing as a constant. Nor does it know a defender was a stride away from blocking, which is why some low-value shots look far better on the replay than on the spreadsheet.
Game state bends the numbers too. A side protecting a 1-0 lead late on takes fewer risks, sits deeper and lets the shot count fall, so its xG for that match understates how well it played. The model also averages across the whole league. A direct team that crosses constantly and a possession side that walks the ball into the box generate the same Team Expected Goals from very different football, and the number flatters one while underselling the other.
What it never touches is the journey: possession value, progression up the pitch, the pass before the pass. Expected Threat (xT) and progressive pass counts cover that ground. Use xG as a summary of chance quality, not a verdict on the performance.
Next time a result lands badly, run the same three checks before you say anything. Look at the xG line first and ask whether the performance deserved the scoreline. Then decide if it was a one-off or part of a run, because a single match tells you almost nothing on its own. Finally, scan for penalties and red cards, since ten men and a spot kick change both the shot count and the context behind it.
If you want to see the metric do real work, pick one team and log its Expected Goals against its actual goals for a full season. Do it in a plain spreadsheet, one row per match. Around the ten-game mark the two columns will look unrelated, and you will be tempted to bin the exercise. Keep going. By May the gap will have narrowed, and you will understand which of those early results were noise.
That is the honest use of the number: context for what already happened. It says nothing about next Saturday.