What Is xG Expected Goals The Core Metric of Modern Football Analytics

The Rise of xG: How Expected Goals Revolutionized Modern Football

After 2014, a strange English abbreviation began appearing frequently in football broadcasts, data analysis, and fan discussions: xG (Expected Goals). It's not a newly invented technology, but rather the core metric of the modern football data revolution. Today, when you open the data panel for any Premier League match, you'll see Manchester City with 18 shots and an xG of 2.4; Liverpool with 12 shots and an xG of 1.8. What do these numbers actually mean? Why can xG reflect a team's true strength better than "number of shots," "possession rate," or "goals scored"? Why do modern elite coaches, data analysts, and sports analytics media regard xG as "the most important data metric"? Let's clarify this core indicator of modern football data.

Why xG Was Needed

Before xG appeared, the main metrics for evaluating a match were goals scored, number of shots, and possession rate.

But these metrics had serious problems:

Goals scored: Most direct, but heavily influenced by luck. One match might see 30 shots result in just 1 goal, while another might see 5 shots produce 3 goals. A single match's goal tally cannot accurately reflect a team's true strength.

Number of shots: More stable than goals, but ignores shot quality. A 20-meter long-range shot and a close-range shot from inside the penalty box are completely different in value. If you only look at the crude "number of shots" statistic, you overvalue poor long-range shooters.

Possession rate: Poorly correlated with shots. Barcelona has long maintained possession rates above 70%, but their goal efficiency is far lower than some counter-attacking teams with only 40% possession. Possession rate is merely "surface-level data."

There was a need for a more precise metric that could simultaneously consider both the quality and quantity of shots. This is the context in which xG was born.

What Is xG?

xG (Expected Goals) translates to "expected goals" in English. Its core logic is:

xG = The sum of "goal probability" for each shot

In other words, each shot is assigned a "goal probability" based on factors like position, angle, shot type, distance, defender proximity, and more. The sum of all shot probabilities throughout the match equals that match's total xG.

Examples:

• A penalty kick: xG ≈ 0.76 (historically, penalty conversion rate is about 76%)
• A close-range shot inside the penalty box: xG ≈ 0.25-0.40
• A long-range shot outside the box: xG ≈ 0.05-0.10
• A header (from a cross in the box): xG ≈ 0.10-0.20

A match's total xG: If a team created 3 penalty box shots (total xG ≈ 0.9) + 5 shots outside the box (total xG ≈ 0.35) + 1 penalty (xG ≈ 0.76), their match xG would be 0.9 + 0.35 + 0.76 = 2.01.

The deeper meaning of xG: This number tells us that based on the quality of chances created in the match, how many goals a team "should have" scored. If actual goals exceed xG, that's "overperformance" (or goalkeeper errors). If actual goals fall short of xG, that's "poor finishing" (or an outstanding goalkeeper).

How xG Is Calculated

xG calculation is based on machine learning models. The specific steps:

Step One: Collect Historical Data

Data analysis companies (like Opta Sports, StatsPerform) have collected millions of shot data points over the past 10-20 years. Each shot records:

• Shot location (x, y coordinates)
• Distance from goal
• Shot angle (relative to goal)
• Shot type (inside foot, outside foot, header, volley, etc.)
• Pre-shot action (direct run, after dribbling, receiving a pass, etc.)
• Defensive pressure (how many defenders nearby)
• Ball state (ground ball, aerial ball, deflection, etc.)
• Shot outcome (goal/no goal)

Step Two: Train Machine Learning Models

Using this historical data, machine learning models are trained (typically random forests, gradient boosting trees, neural networks, etc.). The model learns the probability of scoring given a particular feature combination.

For example:

• "8 meters from goal inside the box, direct shot, no defensive pressure" → 42% goal probability
• "20 meters outside the box, angled shot, marked by defenders" → 3% goal probability
• "Direct free kick from the corner flag area" → 0.5% goal probability

Step Three: Apply to Real Matches

When a shot occurs in a match, the system automatically identifies its features (position, angle, distance, etc.) and uses the trained model to calculate that shot's xG value.

All shot xG values from the entire match are summed to get a team's total match xG.

Specific Applications of xG

Scenario One: Post-Match Assessment of "True Strength"

2022 World Cup Final (France vs Argentina):

• Argentina actual goals: 3
• Argentina xG: approximately 2.1
• France actual goals: 3
• France xG: approximately 1.6

Data interpretation: Argentina created higher-quality shots (higher xG), but penalty shootouts are extremely random. From an xG perspective, Argentina's victory was "deserved"—they had more chances than France.

Scenario Two: Long-Term Team Performance Assessment

2014-15 Liverpool: High goal count, but xG lower than goals scored, meaning their goals relied on overperformance and opponent mistakes rather than inherent strength. Result: 2015-16 Liverpool's performance declined noticeably, because "overperformance" is unsustainable. xG had already predicted this decline.

Scenario Three: Player Ability Assessment

xG can evaluate shooters. For example:

• Haaland: Recent seasons show xG closely matching actual goals, indicating he's a "consistent shooter."
• Mbappé: Actual goals consistently exceed xG—he has "overperformance ability" (possibly due to excellent shooting technique, unique style, or chance-taking ability).
• Lukaku: Actual goals consistently fall short of xG—he has a "chance-wasting ability" (can create opportunities but has poor conversion rates).

Scenario Four: Coach Tactical Assessment

xG tells coaches whether the tactics are "creating chances." If a team has high xG but low goals, tactics are fine but finishing is poor. If xG is low but goals are high, it's luck rather than genuine quality.

Extended xG Metrics

Around xG, the analytics industry has developed several extended metrics:

xA (Expected Assists)

xA = expected assists. Measures the quality of pass chances created. A high-xA player frequently creates high-quality goal-scoring opportunities, even if teammates don't finish them.

xGA (Expected Goals Against)

xGA = expected goals conceded. Measures the quality of chances allowed by a team's defense. Low-xGA teams defend well.

xG per Shot

Measures the "value" of a single shot. Long-range shots have low xG per shot, while penalty box shots have higher values.

xG Chain

Before a goal is scored, the contribution of each player in the attacking sequence. Useful for evaluating defenders' and midfielders' "attacking involvement."

Limitations of xG

Though xG is revolutionary, it has limitations:

Limitation One: Ignores Goalkeeper Performance

xG only calculates shot "goal probability," not the goalkeeper's individual quality. A world-class keeper (like De Gea or Buffon) can save shots that should have gone in, while an average keeper might give up low-xG shots.

Limitation Two: Cannot Reflect Momentum and Psychology

Momentum, psychology, and tactical adjustments significantly impact match results. xG is purely "rational statistics" and cannot capture these irrational factors.

Limitation Three: Data Collection Precision Limits

xG accuracy depends on data collection precision. If data collectors can't see the exact position or defensive pressure from a distance, the calculated xG might be inaccurate.

Limitation Four: Sample Size Requirements

For a single match, xG can fluctuate significantly. xG's true value emerges with "large samples"—like evaluating a season or a player's career.

The Football Revolution xG Brought

xG's popularization has brought several revolutionary changes to football:

Revolution One: More Scientific Club Transfer Decisions

Previously, transfer decisions relied on observation, experience, and intuition. Now, elite clubs (Manchester City, Liverpool, Bayern, RB Leipzig, etc.) have data analyst teams using xG and related metrics to evaluate target players.

Famous case: Liverpool's 2018 acquisition of Mohamed Salah. At the time, Salah's surface stats (goal numbers at Roma) weren't particularly impressive, but his xG per 90 minutes ranked among Europe's top forwards. Liverpool's analytics team convinced Klopp: "This player is undervalued. xG data shows very high potential." After joining Liverpool, Salah's first four seasons averaged elite-level striker goal production, validating xG's predictive power.

Revolution Two: Deeper Match Analysis

Football commentators and analysts now routinely use xG for analysis. Statements like "Manchester City won 1-0, but xG was 2.3 to 0.8, they should have won by a bigger margin" have become mainstream analysis language.

Revolution Three: More Precise Player Training

Coaches can evaluate each player using xG, not just "how many goals." This clarifies whether a low-scoring forward's problem is "missing shots" or "creating too few chances."

Revolution Four: Better Audience Understanding

xG lets casual viewers understand "who should have won this match." Even in a dull scoreline, xG shows who dominated. This enriches the viewing experience.

xG's Future: From xG to More Complex Metrics

The analytics field continues developing xG-related indicators:

1. Post-Shot xG

Calculates "after the shot direction is set, can the keeper save it?" This removes the "shot selection" factor and purely evaluates shooting technique.

2. VAEP (Valuation of Actions)

A broader metric evaluating how each touch (not just shots) affects goal probability, including passes, dribbles, tackles, and all actions.

3. Expected Threat

Measures "by bringing the ball to a specific location, how much does expected goal probability increase?" This provides precise evaluation for non-shooting positions like midfielders and fullbacks.

These metrics together form modern football's "data universe." In the future, team evaluation will be based on dozens or even hundreds of data points, not just "goals scored."

What xG Means for Regular Fans

What can understanding xG bring to ordinary fans?

First: Understand the Match's "Real Direction"

Through xG, you can know "who truly dominated this match," not just who scored. This deepens your understanding of football.

Second: More Objective Star Player Evaluation

Don't just look at "goals." A forward scoring 20 goals with 14 xG versus one scoring 14 goals with 20 xG—the latter is probably "undervalued." This ability makes you a more sophisticated fan.

Third: More Rational Predictions

xG shows a team's "true attacking ability." Combined with other data, you can make more rational match predictions.

xG Changed Football

From widespread adoption around 2014 to today, xG has profoundly changed football's every aspect—from club decisions, coach tactics, player evaluation, media commentary, to viewing experience.

It proves one thing: football isn't "mystical," but a quantifiable, analyzable sport. Those seemingly "magical intuitions" are backed by data.

Yet xG makes us more in awe of the sport, because no matter how precise the data, it cannot fully predict results. That 1% of surprises, that post, that red card—they keep football unpredictably magical.

xG is a window letting us see football's true face more clearly. But it isn't everything. Football's beauty always lies in those incalculable moments.

This is xG: modern football's core data metric, a tool turning us from "watching football" into "understanding football."