What Is Expected Goals (xG)?

Expected Goals, abbreviated as xG, is a probability-based statistical metric that assigns a numerical value to every shot taken in football. This value represents the likelihood that the shot will result in a goal, based on thousands of historical examples with similar characteristics. Rather than treating all shots equally (as traditional statistics do), xG recognizes that a tap-in from two yards and a 35-yard effort are fundamentally different scoring opportunities.

The metric was invented in 2012 by Sam Green at Opta Sports and has evolved from a niche analytical tool into a mainstream metric used by professional clubs, broadcasters, and bettors worldwide. Today, xG is considered one of the most reliable indicators of team and player performance in football.

The xG Scale Explained

Every shot in football receives an xG value between 0.00 and 1.00, where:

• 0.01–0.10 represents a very poor chance (1–10% probability of scoring)
• 0.31–0.60 represents a moderate chance (30–60% probability)
• 0.86–1.00 represents an almost certain goal (86–100% probability)

For example, a penalty kick typically has an xG value of approximately 0.76–0.79, meaning that roughly 76–79% of penalties are converted. A long-range shot from 35 yards with a defender in the way might have an xG of just 0.02, indicating that such shots result in goals only about 2% of the time.

Where Did Expected Goals Come From? The History of xG

The Origins: Sam Green and Opta Sports (2012)

Before the formal invention of xG, football analysis was limited to crude counting metrics: total shots, shots on target, possession percentage. These statistics told you how much a team shot, but not how well they shot. The sport lacked a way to quantify the quality of chances created.

In 2012, Sam Green at Opta Sports developed the first formal expected goals model. Using historical data from hundreds of thousands of shots, Green created a system that could assign a probability value to any shot based on its characteristics. This innovation allowed analysts to answer the question: "How many goals should this team have scored based on the quality of their opportunities?"

The early xG model was relatively simple, primarily considering shot distance and angle. However, the fundamental insight was revolutionary: performance could now be measured not just by results, but by the quality of chances created and conceded.

Evolution of xG Models (2012–Present)

Over the past decade, xG models have become increasingly sophisticated. Different data companies have developed their own proprietary models, including:

• Opta Sports (the original creator)
• Wyscout (owned by Driblab, used by many professional clubs)
• StatsBomb (advanced model incorporating contextual data)
• InStat (Eastern European provider with detailed metrics)

While these models differ slightly in methodology, they correlate highly (~0.95 correlation), meaning they tell broadly similar stories. The refinements over time have included:

1. More granular shot location data — Moving from simple zones to precise distance and angle measurements
2. Contextual variables — Accounting for defensive pressure, goalkeeper position, and assist type
3. Set-piece modeling — Treating corners, free kicks, and throw-ins separately
4. Machine learning integration — Using logistic regression and other algorithms to improve accuracy

Today's xG models are far more sophisticated than the 2012 original, but the core principle remains unchanged: assigning a probability value to every shot based on historical data.

Adoption in Professional Football (2015–Present)

For the first few years after its invention, xG remained a niche metric used primarily by data enthusiasts and forward-thinking analysts. However, adoption accelerated dramatically after 2015. By the 2020–21 season, most Premier League clubs were using xG in recruitment, tactical analysis, and performance evaluation.

A pivotal moment came in 2017 when Liverpool's analytics team convinced manager Jürgen Klopp to sign Mohamed Salah instead of Julian Brandt, based partly on Salah's superior xG values. Salah has since won four Premier League Golden Boots and become one of the greatest signings in Liverpool's history—a validation of xG-based recruitment.

Today, managers like Pep Guardiola, Thomas Tuchel, Eddie Howe, and Thomas Frank openly discuss xG in post-match interviews. The metric has moved from "nerdy stat" to essential analytical tool.

How Does Expected Goals Work? The Mechanics Behind xG

The Core Variables in xG Calculation

Modern xG models consider multiple factors when assigning a probability value to a shot. The most important variables include:

Each of these variables is weighted based on historical data. For example, a shot from 8 yards with no defenders nearby (high xG) is weighted much more heavily than a shot from 25 yards with three defenders in the way (low xG).

The Mathematical Model Behind xG

At its core, xG uses logistic regression—a statistical method that estimates the probability of a binary outcome (goal or no goal) based on input variables. Here's how the process works:

1. Historical Data Collection — Analysts gather data from thousands of past shots (Opta has analyzed nearly 1 million shots)
2. Variable Extraction — For each shot, they extract: distance, angle, body part, assist type, defensive context, etc.
3. Probability Modeling — Using logistic regression, the model learns the relationship between these variables and actual goal outcomes
4. New Shot Prediction — When a new shot occurs, the model compares it to similar historical shots and estimates the probability of a goal

The formula is essentially:

Probability of Goal = 1 / (1 + e^(-z))

Where z is a weighted combination of all the variables mentioned above. The model "learns" the weights from historical data, ensuring that factors with stronger predictive power (like distance) are weighted more heavily than weaker factors.

Different xG Providers and Their Approaches

While all xG models follow the same basic principle, different providers make different methodological choices:

• Opta's model emphasizes distance and angle, with relatively straightforward variable selection
• Wyscout's model includes more contextual variables and is widely used by professional clubs
• StatsBomb's model is known for granular detail and is often considered the most sophisticated
• InStat's model is popular in Eastern Europe and often produces slightly different values

The differences between providers are usually small (typically within 0.05–0.10 xG per shot). What matters most is consistency: using the same provider's data when comparing teams or players over time ensures you're not conflating differences in methodology with actual performance changes.

Why Is Expected Goals Important? Understanding Its Value

Beyond the Scoreline: What xG Reveals

A team can win 1–0 despite having only 0.8 xG, while their opponent had 2.4 xG. Traditional analysis might conclude: "Team A won decisively." But xG tells a different story: "Team A was lucky. Team B created better chances and should have won."

This is xG's core value: it separates luck from skill. Over a single match, luck dominates. A team can score an unlikely goal and concede very little despite poor defending. But over 10–15 matches, luck regresses and underlying quality emerges.

For bettors and analysts, this is crucial. It means:

• A team's recent win streak might be unsustainable if their xG is much lower than their goals
• A team on a losing streak might be due for a rebound if their xG is high relative to their goals
• A player's goal drought might be temporary if their xG is high

Identifying Overperformers and Underperformers

One of xG's most practical applications is identifying when teams or players are performing above or below their expected level:

Overperformers (scoring more than xG suggests) are likely to regress because:

• They're benefiting from unsustainable finishing efficiency
• Their goalkeeper is having an exceptional run
• They're getting lucky with deflections and bounces

Underperformers (scoring less than xG suggests) are likely to improve because:

• Their finishing will normalize
• They're creating good chances that will eventually go in
• Their luck will turn

For example, during the 2024–25 Premier League season, striker Bryan Mbeumo scored 20 goals from just 12.3 xG—a significant overperformance. This caught the attention of Manchester United, who signed him shortly after the season ended, betting that his underlying chance creation (xG) was strong even if his finishing was unsustainably hot.

Similarly, if a striker has 8 goals from 14.5 xG, they're underperforming, suggesting their next 10 matches will likely see them score more goals as regression to the mean occurs.

How Clubs Use xG in Recruitment

Modern football clubs use xG extensively in player recruitment. Rather than just looking at goals scored, they examine:

1. xG per 90 minutes — How many quality chances does the player create or receive per game?
2. Overperformance/underperformance — Is the player's finishing exceptional, or are they benefiting from luck?
3. Positional fit — Does the player's xG profile match the team's tactical needs?

Liverpool's signing of Mohamed Salah in 2017 is the classic example. Salah had excellent xG metrics at Fiorentina and Roma, suggesting he was a quality finisher in a strong position to create chances. His subsequent dominance in the Premier League validated this assessment.

Conversely, clubs now avoid overpaying for players with high goal totals but mediocre xG, recognizing that unsustainable finishing is unlikely to continue.

What Are the Main Variables Affecting xG?

Shot Location: Distance and Angle

Distance and angle are the two most predictive variables in any xG model. The closer a shot is to goal and the more directly on target, the higher the xG value.

• A shot from 6 yards straight on might have xG of 0.35–0.45
• The same shot from 12 yards drops to xG of 0.15–0.25
• From 25 yards, it's down to xG of 0.02–0.05

Angle matters equally. A shot from the goal line (very tight angle) has lower xG than a shot from the center of the box at the same distance, because the shooter has less of the goal to aim at.

Type of Shot and Body Part

Different body parts have different conversion rates:

• Shots with the foot typically have higher xG than headers
• One-on-one situations (just the striker and goalkeeper) have xG values around 0.40–0.60
• Headers from open play are usually lower xG than foot shots from the same position
• Set-piece headers (from corners) sometimes have higher xG than open-play headers, depending on the model

Assist Type and Pattern of Play

How the shot is created affects its xG value:

• A shot from a through ball might have higher xG (the passer threaded it perfectly)
• A shot from a cross might have lower xG (header opportunities are less reliable)
• A shot from a pull-back (ball played back from the byline) typically has high xG
• Counter-attack shots sometimes have higher xG because defenders are out of position
• Set-piece shots (corners, free kicks) are modeled separately with their own conversion rates

Defensive Context and Pressure

Modern xG models increasingly account for defensive pressure:

• A shot with no defenders nearby has higher xG
• A shot with tight marking or defensive blocks has lower xG
• The position of the goalkeeper can affect xG slightly
• Shots after a tackle or interception (high-pressure situations) sometimes have slightly different xG

How Is xG Different from Actual Goals? Overperformance and Underperformance

Expected Goals vs. Actual Goals

In any single match, xG and actual goals can differ dramatically. A team with 1.2 xG might score 3 goals (extreme overperformance), while a team with 3.1 xG might score 0 (extreme underperformance). This is variance—the natural fluctuation that occurs in small samples.

However, over longer periods, xG and actual goals converge. A team that averages 1.8 xG per game will typically score close to 1.8 goals per game over a full season. This is why xG is most useful as a trend indicator rather than a match predictor.

Why Short-Term Variance Exists

Several factors explain why xG and actual goals differ in the short term:

1. Finishing Quality — Some players are naturally better finishers (Cristiano Ronaldo, Robert Lewandowski) and convert higher percentages of their xG
2. Goalkeeper Performance — A world-class goalkeeper (e.g., Alisson, Ederson) might prevent more goals than their xGA suggests
3. Luck — Deflections, rebounds, and bounces create randomness
4. Sample Size — With only 5–10 shots per game, variance is significant

When Does xG Stabilize?

For individual players, xG typically stabilizes after 10–15 games. For teams, a full 38-game season provides a reliable sample. For multi-season trends, 2–3 seasons of data is ideal.

This is why bettors and analysts often wait until 10 games into a season before fully trusting xG data for individual players, but can use team xG after just 5–6 games.

What Are the Limitations of Expected Goals?

Single-Match Context Issues

xG's biggest limitation is that it doesn't account for tactical context within a single match. For example:

• A team leading 3–0 might sit back defensively, resulting in low xG despite being the better team
• A team chasing the game might create many low-quality shots, inflating their xG despite poor performance
• A team might intentionally play long balls, resulting in low xG but effective football

In these cases, xG alone tells an incomplete story. Context matters.

Common Criticisms of xG

Over the years, xG has faced criticism from traditionalists and some managers:

• "It doesn't capture the intangibles" — True; xG is a quantitative metric and can't measure leadership, morale, or momentum
• "My team played better despite lower xG" — Possible in a single match; unlikely over a season
• "xG is too complex" — The concept is simple (0–1 probability scale); the calculation is complex, but users don't need to understand the math

Most modern managers have embraced xG, though they recognize its limitations. Even Pep Guardiola, initially skeptical, now uses xG extensively at Manchester City.

Edge Cases and Model Blind Spots

No xG model is perfect. Some situations that models struggle with:

• World-class finishers (Messi, Ronaldo) who consistently outperform their xG
• Exceptional goalkeepers who prevent more goals than xGA suggests
• Unique player skills (long-range shooting specialists, free-kick takers) that models might undervalue
• Rare tactical situations not well-represented in historical data

These edge cases are why xG should always be combined with other analysis, not used as the sole decision-making tool.

How Can You Use Expected Goals for Betting?

xG as a Predictive Tool

xG is significantly more predictive of future match outcomes than the scoreline of the previous match. A team that won 1–0 despite having 0.6 xG is less likely to win their next match than a team that lost 1–2 despite having 2.8 xG.

Over a full season, a team's xG differential (xG minus xG Against) correlates very strongly (~0.85) with their final league position. This makes xG one of the most reliable metrics for predicting which teams will finish higher in the table.

Finding Betting Value with xG

The key to using xG for betting is identifying value: situations where the bookmaker's odds imply a lower probability than xG data suggests.

For example:

• If a team averages 2.2 xG per game and their opponent allows 1.4 xGA per game, the combined expected goals is 3.6
• If the bookmaker prices Over 2.5 goals at 1.80 (55% implied probability), but xG suggests 70% probability, there's value in backing Over 2.5
• Conversely, if Under 2.5 is priced at 2.20 (45% implied probability) when xG suggests only 30%, there's value in backing Under 2.5

xG Betting Strategies

Here are practical strategies for using xG in betting:

What Are Related xG Metrics? Beyond Basic Expected Goals

Expected Goals Against (xGA)

While xG measures attacking quality, Expected Goals Against (xGA) measures defensive quality. It's the same metric applied to the opponent's shots—how many goals should your team concede based on the chances you allow?

A team with low xGA is well-organized defensively, limiting opponents to difficult, low-probability shots. This is often more predictive of long-term success than clean sheets, because it accounts for goalkeeper performance and luck.

Expected Assists (xA)

Expected Assists (xA) applies the same probability logic to playmakers. Rather than just counting assists (which requires the shot to go in), xA credits a player for creating high-quality chances, even if the striker misses.

A midfielder with 0.8 xA per game is creating excellent opportunities, even if their teammates aren't finishing them. This helps identify underrated creators and overrated players who get lucky with assists.

Post-Shot xG (PSxG)

Post-Shot xG is calculated after the shot is taken, accounting for where the shot was aimed. This is particularly useful for evaluating goalkeeper performance. A goalkeeper with low PSxG Against is preventing more goals than expected, suggesting world-class shot-stopping ability.

Expected Threat (xT)

Expected Threat measures how much a player's action (pass, dribble, tackle) increases their team's probability of scoring. It's more granular than xG and helps evaluate midfielders and defenders whose contributions don't show up in traditional stats.

Common Misconceptions About Expected Goals

Misconception 1: "xG Is Always Right"

Reality: xG is a trend indicator, not a crystal ball. It's highly predictive over 10–15 games but can be wildly wrong in a single match. Variance exists, and luck is real.

Misconception 2: "High xG Guarantees a Win"

Reality: A team with 3.2 xG and their opponent with 0.8 xG will usually win, but not always. Finishing quality, goalkeeper performance, and luck all matter. xG is one factor among several.

Misconception 3: "xG Doesn't Account for Player Quality"

Reality: xG models do account for quality through context. A world-class finisher in the same position as an average striker will have the same xG, but the world-class finisher will convert it more often. This is where the variance between xG and actual goals comes from.

Misconception 4: "xG Is Only for Data Nerds"

Reality: The concept is simple—a 0–1 probability scale. You don't need to understand the mathematics to benefit from xG insights. Anyone can learn to interpret xG data in 10 minutes.

Misconception 5: "xG Replaces Other Analysis"

Reality: xG is best used alongside other metrics (possession, pressing stats, defensive actions) and qualitative analysis (watching the match). It's a powerful tool, but not a complete picture.

Frequently Asked Questions

Q: What does an xG of 0.5 mean?

A: An xG of 0.5 means the shot has a 50% chance of resulting in a goal—a perfectly balanced opportunity. It's neither easy nor difficult; it's a true coin flip.

Q: Is xG the same across all companies?

A: No. Opta, Wyscout, StatsBomb, and InStat use different models and sometimes produce different values (typically within 0.05–0.10). However, they correlate highly (~0.95), so the differences are usually minor. What matters is consistency—use the same provider when comparing teams or players over time.

Q: How long does it take for xG to become reliable?

A: For individual players, 10–15 games minimum. For teams, a full season (38 games) is ideal. Multi-season trends are most reliable for predicting future performance.

Q: Can xG predict match outcomes?

A: xG is predictive but not deterministic. Over 10–15 games, xG correlates strongly with league position. In a single match, luck and finishing still matter significantly.

Q: Why did my team lose despite having higher xG?

A: Several reasons: (1) Poor finishing by your team, (2) Excellent goalkeeper performance by the opponent, (3) Defensive mistakes leading to low-xG goals conceded, (4) Tactical setup (e.g., defending a lead). xG measures chance quality, not results.

Q: Is xG used by professional clubs?

A: Yes, extensively. Most Premier League clubs use xG in recruitment, performance analysis, and tactical planning. Managers like Pep Guardiola, Thomas Tuchel, Eddie Howe, and Thomas Frank have all publicly discussed xG.

Q: What's the difference between xG and xGA?

A: xG measures attacking quality; xGA measures defensive quality. Combined (xG minus xGA), they show a team's expected goal difference and are highly predictive of league position.

Q: Can I use xG to make money betting?

A: Yes, if you identify value. When bookmaker odds imply a lower probability than xG suggests, you have an edge. However, this requires consistent analysis and disciplined bankroll management.

Q: What's post-shot xG?

A: Post-shot xG accounts for where the shot was aimed (not just where it came from). It's more goalkeeper-specific and helps evaluate keeper performance beyond traditional save percentages.

Q: Why do some players consistently outperform their xG?

A: Possible reasons: (1) Exceptional finishing ability, (2) Positioning skill (getting into high-xG situations), (3) Lucky variance (short-term), (4) Weak xG model for their specific shot profile. Long-term outperformance usually indicates genuine skill.

Related Terms

• xG Against
• xA
• Model betting