What Is Expected Goals (xG)? A Plain-English Guide for Match Watchers
When the Result Doesn't Match the Performance
Image: Daily visits on dogballsblog.com from 8/1/06 - 12/20/07, by mil8, CC BY 2.0
Picture a team that loses 1-0 having taken 18 shots to the opponent's four. The scoreboard records the one thing that mattered to the table: a goal. It says nothing about the eighteen attempts that went wide, blocked or straight at the keeper, and it treats a scuffed finish from six yards the same as a hopeful punt from thirty. The final score tells you what happened. It doesn't tell you how likely it was to happen, or how repeatable it is.
Expected goals, or xG, was built to close that gap. Instead of counting shots, it weighs them, giving every attempt a value based on how often chances like it have been scored. A tap-in is worth far more than a shot from outside the box, and the total for a team gives you a rough measure of the quality of chances it created. You'll now see xG sitting next to possession and shots on broadcast scoreboards and match centres. What follows explains where that number comes from and how to read it.
The Quality of Every Chance
Image: It happened again, by Travis Isaacs, CC BY 2.0
The number attached to a single shot is a likelihood, written between 0 and 1. A tap-in from a yard out might carry 0.85; a speculative effort from 30 yards, 0.03. That decimal is the share of comparable attempts that ended in the net, drawn from thousands of historical shots with similar distance, angle and circumstances. It describes the average outcome of that situation, not the fate of the one you just watched.
This is where most confusion starts. A chance worth 0.30 is a good chance, and it still misses seven times in ten. When a striker spurns it, the model was not wrong. The shot landed in the seven.
A team's xG is simply the sum of every shot it took. Twenty efforts at 0.05 each add up to 1.00, the same as a single penalty, which is why a side can pile up shots without creating much danger.
You will find xG in live match centres, on broadcast graphics, on betting sites and in scoreboard apps. A typical display puts the two team names side by side, the score in the middle, and an xG figure beneath each. Some add a shot map, with every attempt marked as a dot sized by its value.
Take a 1-1 draw where Team A finishes on 1.8 xG and Team B on 0.6. The reading is straightforward: A created the better and more frequent chances, B scored from less. It suggests A was the better attacking side on the night. It does not prove A deserved to win, and it says nothing about defending, goalkeeping or the timing of the goals.
Remember that xG is cumulative. It climbs with every shot, so a quiet first hour can turn around in ten minutes. Check the figure late, not at half-time, before drawing conclusions.
How a Shot Becomes a Number
Image: Daily visits on dogballsblog.com from 8/1/06 - 12/20/07, by mil8, CC BY 2.0
The model behind the figure is trained on history. Most public xG systems have learned from databases of historical shots running into the hundreds of thousands, each one tagged with details of the attempt and whether it ended in a goal. Distance from goal and angle to the mouth carry the most weight. So do the body part involved (a header converts far less often than a clean strike with the foot) and how the chance arrived: a pass, a cross, a rebound off the keeper or a scramble. The number of defenders between shooter and goal matters too. Some models add goalkeeper position, or the pressure on the shooter at the moment of contact. From those examples the model learns how often shots of a given description went in, and applies that rate to new attempts. Opta, StatsBomb and understat each build their own version with their own inputs and training sets, which is why the same shot can read 0.12 on one scoreboard and 0.09 on another.
Why Providers Disagree
There is no single official xG. Each provider builds its own model, with its own data, its own features and its own training window, so the numbers will not always match. A shot that reads 0.12 on one site may read 0.09 on another. Neither is wrong; they are answering the same question with slightly different evidence.
The practical answer is to pick one source and stay with it. Compare a team's xG across matches from the same provider and the figures mean something. Mix Opta on Saturday with understat on Wednesday and you are measuring the difference between two models as much as the difference between two performances.
One more thing worth knowing: some providers also publish post-shot xG, which accounts for where the ball ended up inside the goal rather than just where it was struck from. That is a separate metric, useful in its own right, and not a correction to the original figure.
What xG Is Good For
Image: Getting my butt kicked, by juhansonin, CC BY 2.0
The main use is judging whether a result reflected the balance of chances. Take a team that wins 2-0 with 0.7 xG while conceding 1.9. The scoreline says comfortable; the numbers say fortunate. Both are true, and knowing which one you watched changes how you read the next fixture.
Over a run of matches, xG tends to predict future results better than recent scorelines do, because finishing fluctuates. A striker in a purple patch and a side conceding deflected goals will both drift back toward their chance quality. Analysts lean on rolling xG averages across five to ten games for exactly this reason: a single match is noisy, a block of five tells you something about form and approach.
The caution is that xG measures chances, not control. A team can post a healthy figure from three set pieces and still be pinned back for eighty minutes. Defensive structure, pressing and territory are only partly reflected in the number, so read it alongside what your eyes told you.
A striker with 14 goals from 9.5 xG is finishing well. The gap of four and a half goals is the kind of return that gets noticed, and the question worth asking is whether it describes a skill or a hot streak. Sustaining that rate is unusual. Across large samples, most players drift back toward their chance quality, which is why a run like that tends to even out over a season rather than continue indefinitely.
Elite finishers do outperform xG consistently, so the number is not destiny. Harry Kane and Lionel Messi have built careers on converting chances that models rate as difficult, and a handful of goalkeepers do the same in reverse. Expected goals on target, which assesses where shots ended up rather than where they were struck from, gives a fuller picture of keeper performance.
One match proves nothing either way. The same logic that makes overperformance interesting over ten games makes it meaningless over ninety minutes. Treat a gap between goals and xG as a question, not a verdict.
- Compare the two xG totals at full time before drawing conclusions from the scoreline, since a wide gap either way tells you who created the better chances.
- Check the half-time xG figures to see whether a lead was earned or rode on an early goal against the run of play.
- Track a team's rolling xG across several matches to separate a genuine dip in chance creation from a short run of poor finishing.
- Pair xG with shot counts: thirty shots for 1.2 xG means volume without quality, while eight shots for 1.9 means the opposite.
Where xG Falls Short
Image: City Saw Mills Timber and Slate Stores, Dublin, Éire, by wlemay, CC0
Every xG figure is built from shots, and only shots. A side that pins the opponent in their own half for half an hour but never gets a clear sight of goal registers almost nothing, even though the territorial control, the pressing and the patient build-up were real and were what won the match. The same blind spot runs the other way: a well-drilled defence that funnels attackers into harmless positions is credited only with the low xG it conceded, never with the organisation that produced it.
The model also treats every shot as a comparable event. A chance at 0-0 in the 89th minute, with a goalkeeper fully alert and a defender closing, is not the same as the identical chance at 3-0 up when the game is already decided. Nor does the number know who is between the posts. And defensive xG is nothing more than the opponent's xG, which measures the chances faced, not the skill of the back line that faced them.
The most common mistake is reading xG as a forecast. It describes the chances already taken, not the ones to come, so a team with 2.4 xG can lose 3-0 and nothing has gone wrong with the maths: they created enough to score two or three, their finishing was poor, and the opponent converted three of four. That is football, not a contradiction.
Two more traps are worth flagging. Comparing xG figures from different providers is comparing different models, as the earlier disagreement over 0.12 and 0.09 showed. And a high xG does not automatically mean a good performance; it may simply mean a side spent the afternoon lobbing crosses at a defence that dealt with them comfortably.
Be careful with single matches, too. One game's xG is noise. Pair the number with shots on target, possession and pressing data before you draw conclusions, and let a season's worth of matches do the talking.
Read It, Then Watch the Match
Image: Dabr in Indonesia, by Thox, CC BY-SA 2.0
After the whistle, work through it in the same order every time. Note the score. Note the xG totals beneath each team. Then check the shot count and where those shots came from, because twelve efforts from outside the box tell a different story to twelve from inside the six-yard area. If the gap between xG and goals is wide, treat it as a talking point rather than a verdict. Something happened; the number does not tell you what.
The better habit is the longer view. Follow a team's rolling xG across five to ten matches and you start to see whether results are tracking performances or running ahead of them. Used this way, xG works best as a question-generator. Why did a side with 1.9 xG only score once? Who was taking the chances, and from where?
Let the scoreboard's xG display send you back to the highlights with a sharper eye for chance quality. Just remember the caveat: no single number replaces watching the game.