A normal person's guide to expected goals (xG)

We've all seen that match. One team batters the other for ninety minutes, hits the post twice, misses a sitter from three yards, and somehow loses 1-0 to a deflected long-range fluke. You walk away feeling like the scoreline lied to you. That's exactly why expected goals exists. It's a way to put a number on that feeling that one team 'deserved' more.

What does a single xG number actually mean?

An xG value represents the probability of a specific shot being scored, based on thousands of similar shots from the past. If a shot is rated as 0.10 xG, it means that a player would be expected to score that exact chance ten percent of the time, or once every ten attempts.

Think of it as a scale from 0 to 1. A shot from the halfway line might be 0.01 xG—a one-in-a-hundred prayer. A tap-in on the goal line with the keeper out of the picture might be 0.95 xG. When you see a player miss a 'big chance,' you're usually looking at something north of 0.40 xG. It's not about what the player actually did with the ball; it's about how good the opportunity was before they struck it.

This is why you'll see a team finish a game with 2.5 xG despite only scoring once. It suggests they created enough high-quality openings to have scored two or three times, but their finishing was poor or the opposition goalkeeper had the game of his life.

How do the models decide if a chance is good?

Models used by Opta or Understat look at several historical variables to determine how difficult a shot is. The most obvious factor is distance, but the angle to the goal and the type of assist matter just as much.

It's not just about where the boots are on the grass. A modern expected goals explained breakdown usually considers these factors:

Data providers like Stats Perform have mapped out hundreds of thousands of shots to find these patterns. They know that a volley from a cross is harder to control than a square pass across the carpet. By crunching these variables, the model gives us a objective baseline. It removes the bias of our own eyes, which often remember the goal but forget how difficult the chance actually was.

Why should we care about xG more than shots on target?

Expected goals is superior to traditional shot counts because it accounts for the massive disparity in shot quality. Ten weak efforts from 30 yards are not equal to one clear header from the six-yard box, yet a basic stat sheet treats them the same.

If you look at FBref, you'll see that a team might have fifteen shots but an xG of only 0.80. That tells you they were desperate, snatching at chances from distance rather than breaking the defense down. Another team might only have five shots but an xG of 1.50. They are the ones playing the better football, finding the 'high-value' areas.

This leads us to the concept of 'regression.' If a striker is scoring ten goals from 3.0 xG, they are probably on a lucky streak that won't last. Eventually, physics and probability catch up. Conversely, if a world-class striker hasn't scored in five games but is still racking up high xG, you can bet your house they'll start finding the net soon. The process is right, even if the result isn't showing up yet.

What are the limitations of xG?

The biggest thing xG doesn't account for is the identity of the person taking the shot. Most models assume an 'average' player is pulling the trigger.

If Erling Haaland is on the end of a 0.50 xG chance, you'd expect him to score it more often than a League Two centre-back would. However, the model treats the chance the same for both. This is why the very best players in the world often 'outperform' their xG over a full season. They are simply better finishers than the average.

It also doesn't track 'dangerous moments' that don't result in a shot. A cross that zips across the face of goal without anyone touching it is worth 0.00 xG, even though it was terrifying for the defenders. It's a tool for measuring shots, not the entire flow of the match.

You also have to be careful with cumulative xG. A team might have an xG of 2.00, but if that came from twenty tiny 0.10 chances, it's not the same as having two 1.00 chances (which is impossible, but you get the point). Lots of bad shots rarely add up to a win as reliably as a few great ones do.