Expected Goals (xG) Explained

“Expected goals” has become one of the most quoted numbers in football, and it is used for two different things. Broadcasters and analysts mean a measure of chance quality in a match that has already been played. Prediction models, including ours, use the same name for a forecast of how many goals a team will score in a match that hasn’t happened yet. This guide explains both, and how to use the pre-match figure on our match pages.

Shot-based xG: measuring chances after the fact

Shot-based xG gives every shot a probability of becoming a goal, based on what usually happens from similar situations. The inputs typically include:

• distance from goal and the angle to it;
• whether it was a header or a shot with the foot;
• how the chance was created: a through ball, a cross, a rebound, a set piece;
• in more advanced models, the positions of the goalkeeper and defenders.

A tap-in from two yards might be worth 0.9, a shot from 30 yards 0.03, and a penalty usually around 0.75 to 0.8. Add up every shot a team took and you get its xG for the match. A side that lost 1-0 while creating 2.4 xG against 0.6 was, by this measure, unlucky.

The value of shot-based xG is that it is less noisy than goals. Finishing streaks and goalkeeping heroics come and go; the quality of chances a team creates is more stable. Over a run of games a team whose goals far exceed its xG tends to drift back towards it.

Pre-match expected goals: a forecast

The expected goals on a prediction site answer a different question: before kick-off, how many goals is each team likely to score in this match? It is the average of the model’s goal distribution for that side, the starting point from which the correct score grid and every goal market are built.

On GoalStatsHub, the expected goals shown on each match page (for example “xG 1.52-1.28”) are this pre-match figure. Our model estimates them from each team’s scoring and conceding record, home and away performance, recent form and previous meetings. We don’t use shot-by-shot data, so it is not the same number as the xG you might see in a post-match report, and the two shouldn’t be compared directly.

What the two numbers tell you

The pre-match figures carry most of the model’s view of a match in two numbers:

• The total says how open the game should be. Top leagues usually average somewhere between 2.5 and 3 goals a game; a total above 3.2 points to an open match, below 2.2 to a tight one.
• The gap says how one-sided it is. Equal figures mean a coin toss with a sizeable draw chance; a gap of more than one goal usually means a clear favourite.

A few examples from a simple Poisson model show how much the split matters:

Moving the home figure from 1.3 to 2.0 and the away figure from 1.0 to 0.8 raises the home win chance by 22 points. Small changes in expected goals have large effects on the result market, which is why good estimates matter so much.

Using shot-based xG alongside our numbers

If you follow a league closely, post-match xG is a useful cross-check. A team whose results have been much better than its chances may be due a correction that a results-based model is slow to see. A team that keeps creating good chances without scoring may be better than its recent goals suggest. Those are exactly the cases where your own judgement can add something to a model’s figure.

Limits worth remembering

• Pre-match expected goals assume a typical match. A red card, an early goal or a team settling for a draw changes the game in ways no pre-match number captures.
• They are only as good as the data behind them. Early in a season, or when two clubs from different leagues meet in a cup, they rest on few comparable games, and our confidence dots and match notes say so.
• Shot-based xG has its own blind spots: chances that never become shots, such as a cross that just misses a striker’s run, count for nothing.