You're watching a match and the commentator says, "That was a glorious chance — about 0.8 xG." Or you glance at the post-match graphics and see "Expected Goals: 2.3 – 0.6" next to a game that finished 1–1. If you've ever nodded along while quietly wondering what those decimals actually mean, this is for you.
Expected goals (xG) has gone from a niche analyst's tool to something you'll hear on mainstream broadcasts every weekend, especially across Europe. It's not a gimmick — it's one of the most useful ideas in modern football, and once it clicks, you'll never watch a match quite the same way. Here's the whole thing, simply.
What xG Actually Measures
Expected goals answers a single question for every shot:
How likely was this shot to become a goal?
The answer is a number between 0 and 1. A shot with an xG of 0.10 would, historically, be scored about 10% of the time from that situation. A shot with an xG of 0.76 — a typical penalty — gets scored about 76% of the time. Add up all of a team's shots in a match and you get their total xG: roughly how many goals they'd be expected to score given the quality of chances they created.
The whole point is this: not all shots are equal. A tap-in from two yards and a hopeful 35-yard screamer both count as "one shot" in old-fashioned stats — but one is a near-certain goal and the other almost never goes in. xG captures that difference. Here's what it looks like on the pitch:
How Is xG Calculated?
xG isn't guesswork — it's built from data on hundreds of thousands of past shots. Analysts take every historical shot, note its characteristics, and see how often shots like it were actually scored. A model then learns the relationship, so any new shot can be rated against that history.
The main ingredients that decide a shot's xG:
| Factor | Effect on xG |
|---|---|
| Distance from goal | Closer = much higher xG. This is the biggest factor by far |
| Angle to goal | A central position sees more of the goal than a tight angle by the byline |
| Body part | Headers are generally harder to score than shots with the feet |
| Type of chance | A fast counter-attack or a cutback often yields a better chance than a static one |
| Assist type | A through-ball that beats the defense tends to create higher xG than a cross |
Plug a shot's features into the model and out comes its probability. A penalty is always ~0.76 because penalties, historically, go in about three-quarters of the time. A header from the edge of the box might be 0.03. The model is simply the accumulated memory of how often shots like this one find the net.
How big is that memory in practice? Opta — whose model is the one most often quoted on broadcasts — describes an xG model built with a machine-learning method called XGBoost, trained on close to a million historical shots drawn from 40 competitions. Rather than the handful of factors above, it weighs more than 20 context variables per shot, including the goalkeeper's position and whether the shooter was under pressure.
That scale is the whole reason the numbers mean anything. A model trained on a few thousand shots would be guessing; one trained on a million has genuinely seen how often a header from that spot, under that pressure, goes in.
xGOT: The Metric That Fixes xG's Biggest Blind Spot
Here's an honest weakness of plain xG, and the newer metric built to solve it.
xG rates the chance, not the shot. It answers "how good was this opportunity?" — and then stops. It says nothing about whether the player then hit a perfect strike into the top corner or dragged it straight at the keeper. Both get the same xG, because both came from the same position.
Expected goals on target (xGOT) closes that gap. Stats Perform describes it as a logistic-regression model built on hundreds of thousands of on-target shots, combining two things: the shot's original xG and where in the goalmouth the ball actually ended up. A shot placed just inside the post is far harder to save than one drilled at the keeper's midriff, and xGOT prices that in.
That makes it useful for two questions plain xG can't answer:
- Finishing quality. A striker whose xGOT consistently exceeds their xG is genuinely striking the ball better than the average player would from those positions — real skill, not luck.
- Goalkeeping. Compare the xGOT a keeper faced with the goals they actually conceded, and you get a measure of shot-stopping that accounts for how hard each save really was. Conceding fewer than the xGOT faced means saving shots most keepers wouldn't.
If you see xG and xGOT side by side, read them as a pair: xG for the quality of the chance created, xGOT for how well it was struck and how well it was stopped.
Reading a Match With xG
Here's where xG earns its keep: it tells you what the scoreline often hides. Football is a low-scoring sport, which means luck and small margins swing results constantly. A team can dominate, create chance after chance, and still lose 1–0 to a deflected fluke. The scoreboard says they were beaten. xG tells the truer story.
Imagine a match that ends 1–1, but the xG is 2.4 – 0.5. That tells you the team that drew "should," on the balance of chances, have won comfortably — they created 2.4 goals' worth of opportunities and were unlucky (or wasteful) to take only one. The other side scored with almost nothing, riding a hot moment or a brilliant individual strike. The result was a draw; the performance was lopsided.
This is why analysts, coaches, and increasingly fans use xG to look past a single result:
- Did we actually play well, or did we just get a lucky bounce?
- Are we creating good chances, or just shooting from anywhere?
- Is our striker genuinely out of form, or missing good chances that will start going in?
xG Over a Season: Luck vs. Sustainability
Over one match, anything can happen — a single wonder-goal can make xG look silly, and that's fine. xG's real power shows up over many games, where luck evens out.
If a team is consistently scoring more than their xG, they may be enjoying a hot streak (or have a uniquely clinical finisher) — but it often isn't sustainable, and their goals may dry up. If a team is consistently scoring fewer than their xG, they're likely creating good chances and getting unlucky — which usually means better results are coming. The same logic applies to individual strikers: a forward underperforming their xG is often a forward about to score a few.
This is exactly why recruitment departments love xG. A striker on a cold team might be undervalued if their underlying chance quality is excellent — the goals simply haven't arrived yet. xG helps separate genuine quality from short-term variance, which is worth a fortune in the transfer market.
The Limits of xG (Read This Part)
xG is powerful, but it is not gospel, and treating it as the final word is the most common mistake. Keep these caveats in mind:
- It rates the chance, not the strike. Plain xG says nothing about how well the ball was actually hit — that's what xGOT above is for. Two shots from an identical spot get an identical xG whether one flew into the corner or trickled at the keeper.
- Simple models miss context that commercial ones capture. This caveat needs updating: the free public models many fans see are relatively basic, but leading commercial models now do factor in goalkeeper position and shooter pressure. If you're comparing numbers, check whether you're looking at a simple model or a 20-plus-variable one.
- Small samples lie. xG from a single match can be misleading. One penalty inflates it; one screamer that beats the model deflates it. It's most reliable across many games.
- Different providers use different models. The same shot might be 0.09 from one source and 0.12 from another, because they weigh factors differently. Don't treat tiny differences as precise truth.
- It measures chance quality, not entertainment, defending, or game state. A team protecting a lead may "lose" the xG battle on purpose and be perfectly happy.
Used wisely, xG is a flashlight that reveals what the scoreline obscures. Used carelessly — "we had more xG so we deserved to win, full stop" — it becomes a blunt instrument. The number is the start of the conversation, not the end of it.
Where to See xG Yourself
You don't need special access. xG now appears on many TV broadcasts and club graphics, and free public sites like FBref and Understat publish detailed xG numbers and shot maps for major leagues. Pull up your team's last match, look at the shot map, and compare the xG to the final score — it's a genuinely fun new lens on games you've already watched.
Frequently Asked Questions
What does xG mean in football?
Expected goals (xG) is the probability that a given shot results in a goal, on a scale from 0 to 1. A team's total xG estimates how many goals their chances "should" have produced.
How is xG calculated?
A machine-learning model trained on historical shots rates each new shot and outputs the scoring probability for shots like it. Opta's model, the one most often quoted on broadcasts, uses a method called XGBoost trained on close to a million shots across 40 competitions, weighing more than 20 variables per shot — distance and angle above all, plus body part, chance type, assist type, goalkeeper position and whether the shooter was under pressure. The number is essentially the accumulated memory of how often comparable shots have gone in.
What's the difference between xG and xGOT?
xG rates the chance; xGOT rates the strike. Plain xG asks how good the opportunity was and stops there, so a shot smashed into the top corner and one dribbled at the keeper from the same spot score identically. Expected goals on target adds where in the goalmouth the ball actually finished, which makes it a measure of finishing quality — and, viewed from the other end, of goalkeeping. A keeper conceding fewer goals than the xGOT they faced is saving shots most wouldn't.
Why is a penalty always around 0.76 xG?
Because, historically, penalties are scored roughly 76% of the time — so every penalty is assigned that same probability regardless of who's taking it.
Can a team win but lose the xG battle?
Absolutely, and it happens often. Football is low-scoring, so a team can score with few chances and beat an opponent who created far more — winning the match but "losing" on xG.
Is xG a reliable stat?
It's very useful, especially over many matches, but it has real limits: it doesn't capture every detail of a chance, varies by provider, and can mislead in small samples. Treat it as a strong indicator, not absolute truth.
The Bottom Line
Expected goals took a frustrating truth about football — that the best team often doesn't win on the day — and turned it into something measurable. By scoring every chance on how likely it was to be a goal, xG lets you see past lucky bounces and cruel deflections to the performance underneath: who created the better chances, which results were deserved, and which are about to change.
It's not here to replace the joy of watching the game, and it's certainly not the final word on who "should" have won. But the next time a match ends 1–1 and the xG reads 2.4 – 0.5, you'll know exactly what that means — and if there's an xGOT column beside it, you'll know whether the finishing or the goalkeeping was the reason.
Sources
- What is expected goals (xG)? — Opta Analyst, on the model, its training data and variables
- Introducing expected goals on target (xGOT) — Stats Perform
- What are expected goals on target (xGOT)? — Opta Analyst
- Expected goals (xG): the football metric changing analysis, betting and fan engagement — Stats Perform
- Enhancing expected goals on target — Stats Perform, on shot execution and goalkeeper performance
Related on PrimusSource: AI Is Changing Football. The Question Is Whether Football Is Ready for It and How UFC Scoring Really Works — more in our Football topic hub.
