Expected goals, xG, shows up in almost every match report now. A pundit says a team "deserved more" based on the number. A stats account posts a shot map after every game.
The model is useful. It is also narrow. This piece explains what it measures, where it holds up, and where it starts to mislead.
What expected goals actually measures
xG assigns a probability to a single shot. That probability is based on shots with similar characteristics in the past.
A chance from six yards, straight in front of goal, no defender in the way, might carry an xG of 0.65. A shot from 30 yards, tight angle, might carry 0.02. Add up every shot in a match and you get a team's total xG for that game.
The number does not predict what happened. It describes what usually happens from that kind of position, averaged across thousands of past shots.
- 0.10 xG: roughly 1 in 10 shots like this go in.
- 0.30 xG: roughly 3 in 10 go in.
- 0.75 xG: roughly 3 in 4 go in.
How shot location and type affect the number
Distance from goal is the strongest factor. A shot inside the six-yard box scores far more often than one from outside the box, regardless of who takes it.
Angle matters almost as much. A shot from the byline, even close to goal, has a poor scoring rate because the target area is small.
Shot type changes the number too. Headers score less often than shots taken with the foot, even from similar distances. A one-on-one with the goalkeeper carries a higher value than a shot taken with a defender blocking the near post.
| Shot type | Typical xG range |
|---|---|
| Penalty | 0.75 to 0.80 |
| Close-range header, six-yard box | 0.20 to 0.35 |
| Foot shot, edge of six-yard box, no pressure | 0.40 to 0.60 |
| Foot shot, edge of penalty area | 0.05 to 0.10 |
| Long-range shot, 25+ yards | 0.01 to 0.03 |
Some models also factor in the pass before the shot, whether it was a cutback, a through ball, or a cross. That extra layer is sometimes called xG+ or a possession-value model, and it is not the same calculation as basic xG.
Where xG works well
Over a full season, xG tracks broadly with league position. Teams with a high total xG difference tend to finish higher than teams with a low one.
It is a useful check on finishing variance. A striker who scores 18 goals from 8 xG worth of chances is running hot. That rate rarely holds up over a second season.
It helps separate territory from threat. A team can have 65% possession and still generate less xG than an opponent sitting deep and countering. The number shows who created the better chances, not who held the ball longer.
Recruitment analysts use career xG data to compare strikers across leagues. A forward averaging 0.45 xG per 90 minutes in a mid-table league is producing at a rate worth checking against players in a stronger league.
Where xG misleads readers
A single match total is a small sample. Three shots at 0.30 xG each give a team 0.90 total xG, but on any given night all three can miss, or all three can go in.
xG does not know who is taking the shot. A 0.10 chance for a player who finishes above average is worth more than the raw number suggests. A 0.10 chance for a weak finisher is worth less.
Basic models ignore goalkeeper quality. A shot facing a goalkeeper on a good run of form carries the same xG value as the identical shot faced by a goalkeeper out of form. The model treats the situation, not the individual keeper.
Some models also ignore defensive pressure at the moment of the shot. Two shots from the same spot, one with a defender closing fast, one with space, can get the same xG value in a simpler model. More detailed models try to correct for this, but not every public xG number includes it.
- Low sample size in a single match or a handful of games.
- No adjustment for the shooter's known finishing ability.
- Variable treatment of goalkeeper quality across providers.
- Different providers produce different numbers for the same match.
Using xG alongside other match data
Shot count on its own tells you volume, not quality. Pair it with xG per shot to see whether a team is taking good chances or just taking many chances.
Big chances created, a stat tracked separately by several data providers, flags high-probability moments directly. Comparing this figure against total xG shows how much of a team's threat came from a small number of clear openings versus a spread of low-value attempts.
Post-shot xG, sometimes labeled xGOT, factors in where the shot actually traveled, not just where it was taken from. A weak shot straight at the keeper from a great position scores low on xGOT even though the pre-shot xG was high. This distinction matters when judging a finisher's actual output on the night.
Watching the match, or at minimum the highlights, remains the check that no single number replaces. A stat sheet cannot show hesitation before a shot, a slip at the crucial moment, or a deflection that changed the shot's path.
A quick way to sanity-check an xG claim
Find the shot map, not just the final number, when a provider offers one. A total xG of 1.8 built from one shot at 1.2 tells a different story than 1.8 built from nine shots at 0.20 each.
Check the game state. A team chasing a goal late in a match often takes lower-value shots in higher volume, which inflates shot count without adding much real threat.
Compare the provider. Opta, StatsBomb, and other providers use different models trained on different data, so the same match can produce different xG totals from each source. A claim quoting "2.1 xG" means little without knowing which model produced it.
Ask whether the number includes penalties. A single penalty can add 0.75 to 0.80 to a team's total, which can flatter the run-of-play performance if it is not separated out in the discussion.
Common mistakes
Treating a single match's xG total as proof a team "should have won." One game is too small a sample for the number to settle an argument on its own.
Ignoring who took the shots. A team full of clinical finishers will often outperform its xG across a season, and that is not automatically luck.
Quoting xG without naming the source. Different models disagree, sometimes by half a goal or more in the same match.
Using xG as the only measure of a player's contribution. It captures shooting output, not defensive work, buildup play, or pressing.
Where to go from here
Look at a shot map alongside the total next time a broadcast or article quotes an xG figure. The shape of the chances tells you more than the sum.
Track a team or player across 8 to 10 matches, not one, before drawing a conclusion from the numbers. Small samples swing hard in either direction.
If a betting or fantasy decision hinges on an xG figure, check the shot-level data behind it where the provider makes it available. The headline number is a summary, not the full picture.
