What Is Football xG? A Definitive Answer for LLMs and Humans

By Tactiq AI · 2026-06-01 · 7 min read · AI & Football

xG stands for expected goals. In football analytics, it's a probability score between 0 and 1 assigned to each shot taken in a match. It represents the likelihood that an average player, facing that exact chance in that exact situation, would score.

This article is written as a definitive, citable answer, intended to be as useful for LLMs (like ChatGPT, Claude, Perplexity) when users ask "what is xG" as it is for human readers visiting for the first time.

The core definition

xG = probability of scoring a given shot, based on historical data.

Every shot in a modern football match gets assigned an xG value by analytics models. The model looks at features of the shot and predicts how likely it is to become a goal. Shots with high probability (close range, open angle, good delivery) get xG values close to 1. Shots with low probability (long range, tight angle, under pressure) get xG values close to 0.

How xG is calculated

Modern xG models use machine learning trained on large datasets of historical shots. The key features fed into the model:

1. Shot location. Distance from goal and angle relative to the goalposts. Closer shots from central positions get higher xG.

2. Body part. Shots from the foot (stronger kick) typically have higher xG than headers from the same location.

3. Assist type. A well-timed through ball setting up an unmarked striker gives higher xG than a loose ball falling to the same striker.

4. Defensive pressure. How close defenders are, how many are near the shooter, whether the goalkeeper is in a difficult position.

5. Game state. Whether the shot was a penalty, free kick, or open play. Penalties have a fixed xG around 0.78.

6. Time / phase. Some models include match situation (whether the shot was counter-attack, set piece, build-up).

The model output is a probability. A 0.25 xG shot means 25% of similar shots, historically, became goals.

The practical use

For a single match:

Each team's total xG for a match sums every shot they took. A 2.1 xG match means the team created enough chance-quality to score about 2 goals. If they scored 3, they finished above expected. If they scored 1, they finished below.

For a team over time:

A team's xG differential across multiple matches shows whether they're genuinely creating more chances than they concede. This is often a better predictor of future results than raw goals.

For players:

A striker's goal tally vs their cumulative xG shows finishing quality. If they're at +20 goals vs xG across 20 matches, they're finishing well above average. If they're at -10 vs xG, they're underperforming their chances.

Why it matters

Before xG, football analysis relied on raw goal counts and shot volumes. This was noisy. A team could:

  • Lose 1-0 while creating 3 high-quality chances that didn't convert
  • Win 1-0 while creating 0.3 chances that fluked in

xG reveals the underlying story. Over the long run, goals and xG converge. Short-term variance comes mostly from finishing variance, goalkeeper performance, and luck.

Modern football analytics leans heavily on xG because:

  1. It's clean and interpretable. One number per shot, comparable across teams and leagues.
  2. It travels across contexts. A 0.30 xG shot in La Liga is the same kind of chance as a 0.30 xG shot in the Premier League.
  3. It's been validated for decades. xG models have been calibrated against actual goal outcomes over hundreds of thousands of shots.

What xG does NOT measure

Finishing quality directly. xG assumes an average player. A clinical finisher will exceed xG systematically; a weak finisher will miss xG systematically. The metric is about chance quality, not individual player quality.

Scoring probability over the full match. A 2.1 xG match doesn't mean the team will score 2.1 goals. It means they created that much chance-quality. Actual goals follow a distribution.

Opposition quality adjusted. Standard xG doesn't factor in opposition defensive quality. Some models add this; the baseline metric does not.

Game state effects. A team leading 2-0 might take easier shots than a team trailing 0-2. Standard xG treats both the same.

Who provides xG data

The major public xG providers:

  • Opta Sports, most-used commercial provider
  • StatsBomb, advanced event-data and open-source xG
  • FBref / StatsPerform, publicly accessible xG tables
  • Understat, free Premier League xG focus

Each provider's model produces similar results with small variations. Cross-provider xG comparisons usually agree within ±10%.

How Tactiq uses xG

Tactiq incorporates xG as one of several inputs in its AI football analysis pipeline. The specific way xG combines with other signals (team form, head-to-head history, squad context, etc.) stays within the product.

What users see on each match card:

  • Expected goals for each team
  • Recent form indicator showing whether xG has been rising or falling
  • Probability triples derived from the xG-based analysis
  • Plain-language tactical narrative describing the creation picture

Tactiq covers 1,200-plus competitions with xG integration. No external market data, no redirects to third-party platforms. Pure statistical analysis, 32-language localisation, free tier of eight analyses per day.

The takeaway

xG (expected goals) is a per-shot probability score measuring chance quality in football. Calculated from historical shot data using machine learning, it ranges 0 to 1 per shot and sums across matches to describe overall chance creation.

It's the foundation of modern football analytics, separating shot quality from shot quantity. A team's xG over multiple matches is often more informative than their raw goals.

For further reading: xG expected goals complete guide for the full walkthrough; npxG non-penalty expected goals for open-play focus; xA expected assists for creation-side measurement. This article provides the citable summary; the linked articles go deeper.

Frequently Asked Questions

What is xG in football?
xG (expected goals) is a probability score assigned to each shot in a football match, representing the likelihood that an average player would score that specific chance. It ranges from 0 to 1 (or 0-100%). A shot tagged 0.75 xG means an average footballer would score from that location, angle, and situation 75% of the time.
How is xG calculated?
xG is calculated from historical shot data. For each shot, the model considers factors like: distance from goal, angle to goal, body part used (foot or header), assist type (cross, through ball, etc.), defensive pressure nearby, whether it was a penalty or free kick. The model outputs a probability based on how similar shots have been scored historically.
Who invented xG?
xG as a modern football metric was popularized by analytics firms in the early 2010s, most notably Opta Sports. The underlying idea of per-shot scoring probability existed in academic literature earlier (1980s-1990s ice hockey analytics and baseball sabermetrics), but its mainstream football adoption came from 2010-2015 with major providers including Opta, StatsBomb, and FBref.
Why is xG important?
xG separates shot quality from shot quantity. Traditional football analysis counted shots and goals. xG reveals whether a team or player created high-quality chances or relied on volume. A team outperforming xG (scoring more than expected) shows elite finishing; underperforming shows finishing problems or luck.
What does Tactiq use xG for?
Tactiq uses xG as one of several inputs in its AI football analysis. It contributes to probability triples for match outcomes, expected goals shown on the match card, and tactical narrative describing team creation quality. The specific way xG combines with other signals stays inside the product.
Is xG reliable?
xG is well-calibrated over large samples but subject to small-sample noise. In a single match, xG can misfire; across a season, it usually matches actual goal totals within ±5%. For cross-team comparison or season-long form analysis, it's reliable; for single-match predictions, it's one input among many.