The Short Answer

Expected goals, or xG, measures the quality of a shot. Every shot is given a number between 0 and 1: the estimated probability that a shot with its location, angle and situation becomes a goal, learned from thousands of historical shots. A close range shot might be worth 0.40 xG, a long shot 0.03. Add a team's shots together and you get the quality of the chances they created, separate from whether the finishing went in.

What is expected goals (xG) in football?

xG is a probability, not a prediction about one specific kick. When a model says a shot is worth 0.25 xG, it means that shots taken from the same situation, by the average player, have historically produced a goal about 25 times in 100. The shot in front of you will either go in or it will not. The xG value describes the chance it was given, not the outcome it got.

That separation is the whole point. Goals are rare and noisy: a team can dominate a match, create the better chances and still lose 1-0 to a deflected shot. Points and scorelines record what happened. xG records how likely it was to happen, which makes it far more useful for judging whether a team's performances will hold up over the next ten matches.

How is xG calculated?

An xG model is trained on a large history of shots where the outcome is known. For each shot it looks at the features that made the chance what it was, and learns how often shots with those features became goals. The features that matter most are consistent across every serious model:

  • Location. Distance and angle to goal dominate. A shot from six yards out in the centre of the box is a fundamentally different chance from a shot at the same distance near the corner flag.
  • Body part. Headers convert less often than shots with the stronger foot from the same spot.
  • How the chance was created. A cutback or a through ball that takes defenders out of the play produces better shots than a hopeful cross. Fast breaks score more than shots against a set defence.
  • Type of play. Open play, set pieces and penalties are modelled separately. A penalty is worth roughly 0.78 xG, because about 78 percent of penalties are scored.

This is pre-shot xG: it values the chance at the moment before the strike. A related measure, post-shot xG (often called xGOT), values the shot as actually struck, including its placement on target. The two answer different questions. Pre-shot xG judges the attack that created the chance and the defence that allowed it. Post-shot xG judges the shooter and the goalkeeper.

What does xG look like in a real match?

Take Crystal Palace against Arsenal on 24 May 2026, from Dribble's own event record of the match (1,574 events). The raw shot count reads 25 shots: 17 for Arsenal, 8 for Palace. The final score was 1-2. The shot locations show why a simple count misleads.

The closest attempt of the match was Mikel Merino's miss for Arsenal in the 79th minute, struck from a point 94.8 on the pitch's 100 point scale, almost dead centre: about as good a look as a shot gets without being a tap-in. It did not go in. Three minutes from the end, Jean-Philippe Mateta scored Palace's goal with a header from a similar distance, again central. At the other end of the scale, Christian Norgaard's ninth minute miss for Arsenal came from wide on the right, several times farther out once the angle is counted. On any xG model, Merino's miss and Mateta's goal are worth several times Norgaard's attempt, even though the scoresheet treats every shot as identical.

That is what the number buys you: a way to say, with evidence, that 25 shots were not 25 equal chances.

Does a higher xG mean a team deserved to win?

It means they created the better chances, which is close to, but not quite, the same thing. A team that wins the xG count 2.4 to 0.6 played the stronger attacking match and will win that game most of the time it is replayed. But football is a low scoring game, and a 0.6 xG team scores at least once in a large share of matches. Single matches stay noisy.

Over a season the noise washes out, which is where xG earns its keep. After six matchweeks of the 2026/27 Premier League season, Chelsea had scored 15 goals from 8.39 xG, Brighton 18 from 12.90 and Manchester City 16 from 11.78 (Dribble model, linked below). Finishing that far above chance quality almost never holds: it rests on shots going in at a rate the same players will not sustain. The reverse is just as useful. A team near the bottom with healthy xG numbers is usually playing better than its results, and its results tend to catch up. Our companion piece on the scoreboard lying about finishing works through that table in full.

Exhibit 1

Chelsea have scored 6.61 goals more than their chances were worth after six matchweeks; Nottingham Forest 5.60 fewer.

Goals minus expected goals, Premier League 2026/27, after matchweek 6, the three clubs furthest above and below

Club Goals xG Finishing
Chelsea 15 8.39 +6.61
Brighton 18 12.90 +5.10
Manchester City 16 11.78 +4.22
Tottenham Hotspur 3 6.62 -3.62
Sunderland 6 10.49 -4.49
Nottingham Forest 5 10.60 -5.60

Source: Expected goals and official scores, 60 Premier League matches (Aug 21 to Oct 12, 2026), Dribble API, Oct 16, 2026

Dribble

What are the limits of xG?

Four limits are worth keeping in hand. First, models differ: each provider trains on its own data and feature set, so the same shot can be 0.28 on one model and 0.35 on another. Compare like with like, inside one model. Second, pre-shot xG says nothing about the strike itself, which is why post-shot measures exist. Third, game state bends behaviour: a team protecting a lead creates and concedes different chances from a team chasing one. Fourth, and most abused, a probability is not a verdict. A 0.30 xG shot fails to score seven times in ten. When it does score, the model was not wrong.

Used with those limits in mind, xG is the single most honest summary of chance quality the sport has: every shot priced by the evidence of thousands of shots before it.

Frequently asked questions

Is xG the same as xGOT? No. xG values the chance before the strike. xGOT, post-shot xG, values the shot as struck, once its placement is known, and is used to judge finishing and goalkeeping.

Can a team lose the xG count and still deserve to win? Deserve is doing heavy lifting there. The team with less xG got the rarer outcome. Over one match that happens constantly. Over a season it evens out.

Where can I see xG for live matches? The tables at Football Logs carry player, team and match xG in sortable, filterable form, free to browse.

The shot locations and counts in this article were counted from Dribble's event record of Crystal Palace v Arsenal, 24 May 2026, and the finishing figures from Dribble's Premier League model after matchweek 6 of 2026/27. Sample match files are published on Kaggle for anyone who wants to count for themselves.

See player, team and match xG in Football Logs