xG vs xA: How Expected Goals and Expected Assists Differ

Expected goals (xG) measures the quality of a shot, while expected assists (xA) measures the quality of a pass that creates a shot. Both numbers are produced by the same family of statistical models, but they describe different moments in an attacking sequence: xG is about finishing chances, xA is about manufacturing them. Understanding the gap between the two is one of the cleanest ways modern football fans separate strikers who score from playmakers who create, and it is also why a midfielder with no goals can still have one of the highest xA totals in a league.

What xG actually measures

xG is a probability. For every shot taken in a match, a statistical model assigns a value between 0 and 1 representing the chance that an average player would score from that position. A tap-in from two yards out might be worth 0.85 xG. A speculative effort from outside the box might be 0.03 xG. The total xG for a team in a match is simply the sum of these shot values.

The inputs that drive an xG model typically include the distance to goal, the angle to goal, the body part used, the type of pass that preceded the shot, the defensive pressure on the shooter, and whether the chance came from open play, a set piece, a penalty, or a counter-attack. Different providers, including Opta, StatsBomb, and Understat, train their own models on different historical datasets, so the exact xG figure for a single shot will vary slightly depending on the source.

What xG does not measure is the identity of the shooter. A penalty taken by a world-class striker still scores around 0.78 xG, the same as a penalty taken by a defender, because the model is judging the situation rather than the person finishing it. Player-specific finishing skill is layered on top through metrics such as goals minus xG (or G-xG), which captures whether a player tends to over- or underperform expectation.

What xA actually measures

xA goes one step earlier in the chain. Instead of asking how likely a shot is to be scored, it asks how likely the pass that set up the shot was to produce a goal. The model takes each completed pass that led directly to a shot, calculates the xG of the resulting shot, and credits that probability back to the passer as expected assist value.

A through-ball that releases a striker clean on goal might generate a shot worth 0.45 xG. The passer earns 0.45 xA, regardless of whether the striker buries the chance or skies it. A sideways pass that ends with a hopeful 30-yard volley might earn the passer just 0.02 xA. Add up every key pass a player makes over a season and you have their xA total.

Crucially, xA only counts the final pass before the shot. The defence-splitting carry that created the opening, or the switch of play three passes earlier, is not credited. To capture those earlier contributions, analysts use companion metrics such as expected threat (xT), buildup xG, or "shot-creating actions" — the two passes or carries that immediately precede a shot. These metrics live next to xA rather than replacing it.

The key differences between xG and xA

The cleanest way to hold the two metrics in mind is to remember that xG belongs to the shooter and xA belongs to the passer. Beyond that, a few practical differences emerge once you start reading match data with both numbers:

Scope of the chain. xG covers one event (the shot). xA covers two events (the pass and the shot). xA only exists when a pass actually leads to a shot — a clever pass that never produces a shot earns zero xA, even if it was the best ball played all match.
Volume vs concentration. Striker xG totals tend to be smooth and high-volume because forwards take many shots per match. xA totals are spikier — a single defence-splitting pass can carry the entire xA figure for a creative midfielder on a given day.
Correlation with goals and assists. Across a full season, xG correlates very strongly with actual goals for most players. xA correlates with actual assists, but more loosely, because the finisher's quality and luck both sit between the pass and the assist getting registered.
Who tops each chart. Strikers and forwards dominate xG tables. Number 10s, deep-lying playmakers, full-backs with crossing duties, and set-piece takers tend to dominate xA tables. Players such as Kevin De Bruyne and Bruno Fernandes have historically combined elite xA totals with above-average xG totals — a rare double profile.
Use in scouting. xG is often used to evaluate finishing efficiency and shot selection. xA is more commonly used to evaluate creativity and final-third decision-making.

Why the difference matters when reading a match

A scoreline of 2-1 hides almost everything that happened in 90 minutes. Add xG and xA, and a far richer picture appears. Imagine a match where the home side scores twice from 0.6 total xG, while the away side loses despite producing 2.3 total xG. The expected goals tells you the away side created the better chances and probably deserved more from the game. Now layer in xA. If the away team's xA was concentrated in two players, you know who actually manufactured the danger; if it was spread across the whole midfield, you know the buildup was collective.

Live football data platforms increasingly publish both numbers together for the same reason. Platforms such as RubiScore track shot-by-shot xG and pass-by-pass xA across major leagues, which lets readers see not only that a team is performing above or below their goal total, but also which individuals are doing the creating. A striker with high xG but low conversion is often a finishing-luck story that regresses. A midfielder with high xA but few assists is often a teammate-finishing story — the passer is doing their job; the strikers are missing.

Common misreadings to avoid

xG and xA are powerful, but they are routinely misread when fans first encounter them. A few traps come up repeatedly:

Single-match obsession. xG and xA are designed to be read across many matches. A 90-minute sample is too noisy to draw strong conclusions about a player's quality.
Treating xA as "creative talent" outright. A player with no xA is not necessarily uncreative — they may operate deeper in the pitch, where their progressive passes generate shots two passes later, beyond xA's scope.
Ignoring penalties. Penalties carry roughly 0.78 xG each, so a player with two penalty conversions can look like an xG overperformer when the underlying open-play number tells the opposite story. Many analysts now publish "non-penalty xG" (npxG) alongside the raw number.
Comparing across providers. Opta's xG model is not StatsBomb's, which is not Understat's. The directional story usually agrees, but the absolute numbers do not.
Reading xA as "would have been an assist." xA only requires that the pass led to a shot, not that the shot was on target. A pinpoint cross headed wide still earns its xA.

How xG and xA complement each other

The most useful applications of xG and xA happen when the two numbers are read side by side. A striker's xG tells you whether their volume of chances justifies their goal total. The xA of their teammates tells you whether the chances arrived from good service or from the striker manufacturing their own shots. When the team's xG climbs but no individual xA stands out, the side is creating through structure. When one player's xA dominates, the team's chance creation runs through that single conduit — useful information for the opposing coach.

For modelling purposes, xG and xA also feed into more advanced composite metrics. Expected goals plus expected assists (xG+xA) per 90 minutes is a popular shorthand for total attacking contribution, often normalised by position. Possession-value frameworks such as xT (expected threat) and VAEP (Valuing Actions by Estimating Probabilities) build on the same principles, extending the credit beyond just the shot and the pass before it.

Reading both numbers like an analyst

The simplest discipline for reading xG and xA together is to ask two questions of every match. First: did the chances created reflect the scoreline? That is the xG question. Second: who built those chances? That is the xA question. A 2-0 win where the winning side outscored their xG by 1.4 was probably fortunate, especially if the losing side's xA was high and scattered across multiple creative players. A 1-1 draw where one side dominated xG and xA is a controlled performance that did not produce a result.

Football data platforms such as rubiscore.com present both metrics live during matches, alongside the running scoreline. Pairing the two — not just reading one in isolation — is what turns expected goals and expected assists from buzzwords into a working language for understanding how a match was actually played.

My Financial Independence Journey