There’s a big difference between probability and prediction, but you wouldn’t know it from the way people talk about xG. It’s become a modern footballing punchline — misunderstood, misquoted, and increasingly misrepresented.
So what is it?
Expected Goals (xG) is a retrospective measure. It tells you how likely a shot was to result in a goal, based on thousands of similar chances in the past. A 0.3 xG chance means that, historically, players score from that kind of position about 30% of the time.
A penalty is around 0.75. A wild hit from 30 yards? That’s usually about 0.01.
But here’s the key: xG doesn’t have to map to the final scoreline. It simply says, “Based on the quality of chances you had, teams usually score around this many goals.”
That’s not a forecast. It’s context.
Probability ≠ Prediction
Annoyingly, people continuously conflate the two. They see a team rack up 2.9 xG and lose 1-0, and they treat the stat like it’s failed. But xG isn’t trying to tell the future or justify a scoreline. It’s just saying that, more often than not, chances like those result in goals.
It seems to wind a lot of people up, but it doesn’t promise anything. Sometimes that correlates nicely and everyone spots the pattern. But most of the time, football does what it wants, and even though that’s how it’s always been, people seem to hate when it does that and the numbers don’t explain it.
The myth of deserving goals
There’s a really baffling narrative that creeps in when a team scores brilliant goals from low xG chances, that somehow, those goals or the scoreline were an unfair reflection of the match.
Not to get too flowery about it, but are we saying that beautiful goals don’t count? Or that to deserve a result you have to create clear-cut chances to ensure your xG lines up? Almost as if football is supposed to be a game of tap-ins and penalties.
But let me hit you with some knowledge. Scoring from 20 yards is hard – that’s why it carries a low xG. That doesn’t mean you didn’t deserve it or it was a fluke. It means you executed a difficult thing brilliantly. And that should be celebrated, not discounted.
xG isn’t punishing great goals or good finishes; it’s just recognising their rarity. And sometimes in a game the only goals you score are good ones. Overperforming your xG isn’t a bad thing and certainly not the only measure to assess a team’s dominance or the deservedness of a scoreline.
If anything, the more telling stat is when teams underperform their xG. A team that racks up 4.0 xG but doesn’t score has likely created plenty of good quality chances and either finished poorly, been denied by outstanding goalkeeping, or just missed those chances, because – you know – players miss chances.
But it’s still worth knowing. It says something about the process and will help you assess.
Yes, it has flaws
To be fair, some of the frustration with xG isn’t completely unfounded. And I think we saw this with the response to the Coventry result on Saturday.
Sometimes you watch a game, see brilliant chances tucked away nicely, and then head to Twitter and see people getting very confused about there only being a 1.26xG.
That disconnect from the final result does feel strange. It did to me when I saw it. It was a clinical, dominant display, especially in the first half. Within ten minutes WhatsApp group chats were flush with early predictions of “this could be a million nil”, and it went pretty close to that. Those watching could sense the danger and the errors QPR were making and how well Coventry were exploiting those.
Realistically, three of the five goals in the first half felt very, very “score-y”, but the xG didn’t seem to match up.
Goal 1 | Haji Wright | xG: 0.10
Haji Wright flying onto a perfect ball full pelt with the defence off balance? You’d back him. However, history has that chance as 0.10 according to xG.
Goal 2 | Brandon Thomas-Asante | xG: 0.27
BTA had all the time in the world following a defensive mistake and a ball rolled perfectly into his path, with just the keeper to beat. In the real world, 2-0 was a fair outcome based on seeing both these chances play out, but the xG has the opening goals at a combined xG of 0.37.
Obviously that feels a little off. But when that happens, I like to think of the alternative. Imagine a scenario where Haji Wright sent his first chance wide, or the keeper saved BTA’s effort? Frame things like that and suddenly you see an eventuality that feels just as feasible.
Goal 4 | Haji Wright | xG: 0.28
Another chance that felt likely to be scored, and was, was Haji’s second which took the score to 4-0. Caught again on the break, BTA laid it into Wright’s path and he wrong footed flailing defenders and a keeper. Feels low on the face of it, but there was work to do and plenty of reasons why that chance may have been missed. Again – xG is factoring that in.
So that’s the thing: xG models aren’t perfect and they’re not trying to be. Most of them don’t (and can’t) capture every layer of context that help us as fans understand any goal in any given moment.
Because they’re comparing against historical (not individual player) data, they don’t take into account who the striker is. They don’t know if the keeper slipped. They don’t always register the speed of play, a clever dummy, or the fact that a goal just feels inevitable from the moment the assist was played. They’re built to evaluate types of chances, not specific situations in all their nuance.
But that doesn’t make the metric useless; it just means they have limitations. And those limitations are worth acknowledging.
The model can undersell the danger. And when it does, it’s not crazy to question it, but don’t throw the whole thing out because you’re expecting a perfect correlation. Because over time, especially across a season and at larger volume, xG still does a remarkably good job of separating consistent performance and creativity from game-by-game randomness.
The real problem? Interpretation
Ultimately, xG is not a moral compass. It’s not here to tell you who should have won based on fairness in the cosmos. It’s merely a tool for understanding the flow and quality of a game’s chances, beyond the final score. Because it doesn’t perfectly align with the end result doesn’t mean there’s a problem with the stat. It’s how it gets framed – often as a justification or a complaint, rather than as the contextual lens it was meant to be.
Sometimes teams score four goals from nowhere. Sometimes they miss everything from six yards. Sometimes a chance feels easy based on confidence, but historically it hasn’t been. That’s the game and that’s the data. xG just gives us another way to understand it.
Where it gets lost is in the desperation of football fans for everything to be neatly tied up in an easy-to-understand, definitive pattern.
xG is not a replacement for intuition or enjoyment. It’s just a number, one that can help explain performance, but never dictate or predict it.
So stop getting it wrong.
