Football Statistics Explained: xG, xA, Possession and What They Measure
Modern football coverage arrives with a box of numbers: expected goals (xG), expected assists (xA), possession, shots, pressing figures and more. Each one measures something narrower than its name suggests. Expected goals estimates the quality of chances, not the score a team deserved. Possession records who had the ball, not who was in control. Knowing what sits behind each figure makes match reports, previews and betting markets far easier to read.
Expected goals (xG)
Expected goals gives every shot a value between 0 and 1: the estimated probability that a shot of that type, from that position, ends up in the net. The estimate comes from a model built on large numbers of past shots with similar features. Typical inputs include:
- distance from goal and the angle to it;
- the body part used, foot or head;
- how the chance was created, such as a through ball, a cross or a rebound;
- the situation: open play, set piece, counter-attack or penalty.
A tap-in from a yard out carries a high value; a speculative effort from 35 yards a very low one. Penalties carry a fixed, high value in most models. Adding up a team's shot values gives its xG for the match.
What xG is good for
Because goals are scarce, results are noisy. xG cuts through some of that noise by focusing on how many good chances a team creates and allows. Over a long run of matches, large gaps between a team's goals and its xG often narrow, which is why analysts use the measure to judge whether a run of results looks sustainable.
Where xG falls short
- Models differ. Providers use different inputs, so two sources can give the same match different xG figures.
- It only sees shots. A brilliant move that ends without an attempt adds nothing.
- It treats finishers as average. Elite strikers and goalkeepers can beat, or fall short of, the model for long spells.
- Game state distorts it. A team leading comfortably may sit back and allow low-quality chances it does not mind conceding.
Expected assists (xA) and related measures
Expected assists credits the passer. Each pass that leads to a shot is given the xG value of that shot, whether or not the shooter scores. A creative midfielder whose teammates keep missing will show a high xA and few actual assists. Close relatives include expected goals against (xGA), the quality of chances a team concedes, and non-penalty xG, which strips out penalties to give a cleaner picture of open-play chance creation.
A quick reference table
| Statistic | What it measures | Common misreading |
|---|---|---|
| Shots | All attempts at goal | More shots always means more danger |
| Shots on target | Attempts that would have gone in without a save | Every on-target shot is equally dangerous |
| xG | Quality of chances created | The score a team "should" have had |
| xGA | Quality of chances conceded | A pure measure of the goalkeeper |
| xA | Quality of chances set up by a pass | The same thing as actual assists |
| Possession | Share of the ball, by passes or time | Proof of dominance |
| PPDA | Opponent passes allowed per defensive action; lower suggests more aggressive pressing | A verdict on how well a team presses |
For the definitions behind on-target attempts, see our guide to what counts as a shot on target.
Possession: the most quoted, least explained number
Possession is calculated from passes or from time on the ball, depending on the provider. A side can enjoy a large share of the ball and still struggle to create chances, especially against opponents who are happy to defend deep and break quickly. That is why many analysts look at where possession happens, in the final third or deep in a team's own half, rather than at the headline percentage.
Using stats sensibly around betting
Statistics describe what has happened. They do not hand anyone an advantage by themselves, for two reasons. First, the same numbers are freely available to operators, who build them into their prices. Second, small samples mislead: a handful of matches can produce xG figures that look decisive and later turn out to be noise. Combining stats with team news and wider context, as discussed in our piece on head-to-head records and form guides, gives a fuller picture. The role of live data feeds on betting platforms is covered separately in our look at real-time data on Toto sites.
Is xG a prediction?
No. It describes chances that already happened. Some models project future performance from it, but that is a separate step built on further assumptions.
Why does one source show 1.8 xG and another 2.1 for the same match?
Providers train their models on different data and include different features. Neither number is wrong; they are different estimates of the same thing.
Numbers make football richer to follow. If they ever nudge you towards a bet, remember that betting is for adults only, that it is a pastime with a cost rather than a way to earn, and that local rules decide whether it is allowed at all.
