A team loses 1-0 and the graphic underneath says "xG 2.4 - 0.6". Pundits call it robbery, fans of the winning side call xG nonsense, and everyone moves on without agreeing on what the number means.
xG is not complicated. Once you see how one shot gets its value, the rest follows, and so do the limits.
The Short Definition
Expected goals (xG) is the probability that a shot becomes a goal, based on how often similar shots have been scored in the past.
Every shot gets a number between 0 and 1:
- A tap-in from two yards might be worth 0.8 or more.
- A penalty is usually worth around 0.75 to 0.8.
- A header from the edge of the six yard box might be 0.2.
- A shot from 30 yards is often 0.03 or less.
Add up every shot a team takes in a match and you get its xG for that game. A team with 2.4 xG created chances that, on average, would produce a little more than two goals.
How a Single Shot Gets Its Value
An xG model is trained on a large history of shots, often several hundred thousand. For each one it knows the outcome (goal or no goal) and a list of details about the situation. The model learns how each detail changes the chance of scoring.
The details that matter most:
Simple public models only use the first few rows. Commercial providers add defender positions, the keeper location and even the height of the ball. That is why two websites can show different xG for the same match. Neither is wrong; they are using different models.
Reading xG for a Team
A single match xG is a story about chance quality, not a verdict. The 2.4 to 0.6 game above tells you the losing side created far better chances. It does not tell you it deserved to win, because football is decided by goals, and a team can create good chances and still lose.
Where xG becomes really useful is over many matches:
- xG for: how many good chances a team creates.
- xG against (xGA): how many good chances it allows.
- xG difference: the gap between the two, usually a better guide to future results than goal difference or points after ten or more games.
A side sitting tenth in the table but near the top of the xG difference rankings has usually been unlucky, and tends to climb. A side high in the table with a poor xG difference has usually been riding hot finishing or an inspired goalkeeper, and tends to drop. That is the core reason prediction models and league simulators lean on it.
Reading xG for a Player
For strikers, xG answers a fair question: is this player getting into good positions? A forward with 0.6 xG per 90 minutes is consistently getting chances, even during a run without goals.
Comparing goals to xG is tempting, but read it with care:
- Short term gaps are mostly noise. A striker ten games into a season who is three goals above their xG is not necessarily an elite finisher.
- Long term gaps can be real. Over several seasons, a small number of players do beat the model consistently, usually because they pick better shooting moments than the model can see.
- Use npxG for strikers. Penalties inflate the numbers of whoever takes them, so non-penalty xG gives a fairer comparison.
Common Mistakes When People Quote xG
1. **Treating one match as proof.** xG is a probability. A team with 2.0 xG still scores zero goals in a meaningful share of games.
2. **Adding up chances that cannot both happen.** If a shot is saved and the rebound is scored, both shots count toward xG. Some models cap a sequence so it does not exceed what was really possible.
3. **Ignoring game state.** A team leading 2-0 often sits deep and concedes a lot of low quality shots. Its xGA rises without it being in real danger.
4. **Comparing numbers from different providers.** Models differ, so compare like with like.
5. **Forgetting that xG ignores the finish itself.** A shot struck into the corner and one hit straight at the keeper have the same xG. Post-shot xG (sometimes called xGOT) is a separate measure that accounts for where the shot went.
Why xG Powers Football Simulators
If you want to know how a league will finish, you need a rating for every team that is not distorted by a lucky run. Points are noisy. Goals are less noisy. xG is less noisy still, because it counts every chance rather than only the ones that went in.
That is why most title odds models blend real results with xG before simulating the rest of the season. Our explainer on how football title odds are calculated walks through that process step by step, from team ratings to 10,000 simulated seasons.
See xG at Work in Simulig
Simulig uses exactly this idea for the Turkish Super Lig. Team strength is calculated from real match results and xG, updated every week, and then the remaining fixtures are played 10,000 times to give each club a title, Europe and relegation probability. You can also set the score of any upcoming match with "What if?" and watch the table recalculate.
Two honest notes. The model only knows what xG and results can tell it, so it will not see a key injury or a new manager until those show up in the numbers. And Simulig is in Turkish and currently available only on the Turkish App Store. If you follow the Super Lig and read Turkish, it is a free download, with an optional Pro subscription for the full career mode, scenarios and tournaments. It is for entertainment, not betting advice.
The Short Version
xG puts a value on every shot based on how often similar shots go in. Summed over a match it describes chance quality; summed over a season it is one of the best public guides to how good a team really is. It does not decide who deserved to win a single game, and it does not measure the finish itself. Used over enough matches, it cuts through a lot of luck.