How to Read xA, xGOT, and Progressive Passes: Advanced Football Stats Part 2 – Passing Creativity and Post-Shot QualityGuides · Football · Stats explained

How to Read xA, xGOT, and Progressive Passes: Advanced Football Stats Part 2 – Passing Creativity and Post-Shot Quality

Key points

  • xA measures passing creativity, xGOT measures shot placement
  • Progressive pass definitions vary by provider and cannot be compared across platforms
  • Assists are the outcome, xA is the process

1Summary

xG answers the question of "how good was the chance for this shot." But football is not just about shooting; it also includes "who delivered the ball to the shooter" and "what happened after the shot." This article covers three common advanced metrics:

MetricQuestion AnsweredWho It Evaluates
xA (Expected Assists)What is the probability that this pass becomes an assist?The passer
xGOT (Expected Goals on Target)Where did this shot on target land, and how much is it worth?Shooters, goalkeepers
Progressive passesDid this pass move the ball noticeably toward the goal?The passer, the team's progression method

My view is: xA is most valuable for evaluating creative players, xGOT is best for assessing goalkeepers and shooters, and progressive passes are least suited for grading. The first two have publicly released official model explanations, while the latter has different definitions across providers, meaning figures cannot be compared across platforms.

If you have not yet read the foundational guides, it is recommended to first read How to Read Football Stats: Possession, Shots, Shots on Target, and PPDA and What Is xG?, as this article directly builds on concepts from them.

2xA: Converting "Passing" into Probability

How It Is Officially Defined

Opta (the data provider under Stats Perform) defines xA as: "measures the likelihood that a given completed pass will become a goal assist. It considers several factors including the type of pass, pass end-point and length of pass" (Source 1). Its values fall between 0 and 1, where 0 indicates that such a pass is virtually impossible to become an assist, and 1 means the recipient should score every time. The model is built using logistic regression, and its primary variables include pass type (cross, through ball, header, etc.), game state/phase (open play, corner, free kick, throw-in), receiving location, pass origin location, and pass distance (Source 2).

The key point is: xA evaluates the pass itself, without considering whether the recipient ultimately takes a shot or scores.

Why Not Just Look at Assist Totals

An assist only counts when "a teammate scores after the pass," which effectively credits or penalizes you for your teammate's finishing. The example cited by Opta is Liverpool's two full-backs in the 2019-20 Premier League (Source 2):

PlayerOpen-Play xAActual Open-Play Assists
Trent Alexander-Arnold7.16
Andy Robertson4.910

(According to Opta's original text. Both xA and assists in the table count only open play; their full-season total assists including set pieces were 13 and 12, respectively.)

Alexander-Arnold delivered higher-quality balls, but because his teammates failed to capitalize, he ended up with fewer assists than expected; Robertson, on the other hand, had teammates who finished more of his chances. Looking only at assists, Robertson was well ahead; looking at xA, the answer to whose passing was more dangerous flips.

How xA Differs from "xG from Key Passes"

Many websites also sum up the "xG of the shot that followed a key pass" as a metric for chance creation. Opta itself notes two issues with this approach: it relies on the receiver actually taking a shot, and it fails to account for what the receiver does between receiving the pass and shooting (Source 2). In contrast, xA assigns a value to every successful pass, regardless of whether a shot ultimately occurs.

How to Read It

Suppose two midfielders over a season (the following are illustrative numbers, not real players):

AssistsxA
Player A48.0
Player B95.0

Player A created 1.6 times the chance volume of Player B, yet teammates converted only 4 goals; Player B gained 4 extra assists thanks to teammates' hot finishing. If the squad remains unchanged next season, Player B's assists will likely regress toward 5, while Player A's have a chance to rebound. This follows the same reasoning as using Expected Points (xPts) to distinguish team luck.

Limitations

  • An xA value only tells you "the location and manner of the pass," without factoring in teammate positioning or defensive pressure (the variables in Sources 1 and 2 do not mention defender positions).
  • Set-piece xA (corners, free kicks) is typically concentrated among a few designated takers, so totals should be separated into "open play" and "set pieces."
  • In small samples, a single pass with high xA can inflate the entire season's numbers, so at least a dozen or so matches should be examined.

3xGOT: After the Shot, Placement Is What Matters

How It Is Officially Defined

Opta explains: xGOT (expected goals on target) measures "the likelihood of an on-target shot resulting in a goal based on the combination of the underlying chance quality (xG) and the end goal location of the shot" (Source 3). It awards higher values to shots in the corners of the goal and lower values to shots directed straight down the middle at the keeper; values likewise range from 0 to 1 and apply only to shots on target, as off-target shots cannot result in a goal and receive no valuation (Source 3). Opta's model is built on logistic regression using hundreds of thousands of historical on-target shots (Source 3).

Hudl StatsBomb's equivalent metric is called post-shot xG (PSxG), which is also evaluated after the shot is taken and primarily used to assess goalkeeper shot-stopping ability (Source 4). The inputs for the two are not entirely identical: Opta's xGOT considers the original xG plus the end coordinates of the shot within the goal frame (Source 3); StatsBomb's explanation states that PSxG typically looks at shot placement, with some models also incorporating goalkeeper positioning or shot speed (Sources 4 and 7).

Thus, xG and xGOT represent numbers at two different points in time: xG is pre-shot, assessing only the chance; xGOT is post-shot, additionally evaluating the quality of the execution.

Evaluating Shooters: Clinical Finishing or Just Good Luck?

Opta's example is Harry Kane in the 2019-20 Premier League (Source 3). Opta's original text notes that these figures exclude penalties:

Non-Penalty xGNon-Penalty xGOTNon-Penalty Goals
10.914.516

His xGOT is significantly higher than his xG, showing that his shots were not just taken from good positions, but placed in areas that are harder to save; his 16 goals are slightly higher than his xGOT, with the remaining gap attributable to luck or goalkeeper errors.

A simple step-by-step reading sequence:

  1. Goals ≈ xG: Scoring matches chance volume; normal performance.
  2. Goals > xG, and xGOT > xG: Shot placement is genuinely strong; there may be genuine finishing ability.
  3. Goals > xG, but xGOT ≈ xG: Ordinary placement that still went in; more likely luck, prone to regression.

(This three-step framework is my own summary, not an official statement.)

Evaluating Goalkeepers: How Many "Should-Be Goals" Were Stopped

Subtracting actual goals conceded from the xGOT faced by a goalkeeper reveals how many more goals they prevented compared to average. Opta's example is Dean Henderson in 2019-20: facing an xGOT of 39.4, he conceded only 32 goals (Opta's original text notes this excludes penalties and own goals), preventing roughly 7.4 additional goals (Source 3). If a goalkeeper faces an xGOT of 10.0 but concedes 12 goals, they stopped 2.0 fewer goals, performing below average.

This is fairer than looking solely at "goals conceded" or "save percentage": facing 10 highly difficult shots versus 10 tame shots hit straight at the body will result in vastly different save percentages, and xGOT accounts for this difference.

Limitations

  • Applies only to shots on target. Off-target or blocked shots have no xGOT, so it cannot be used to evaluate "chance creation ability," which requires looking back at xG or xA.
  • A shooter's finishing ability is mingled with model noise: do not jump to conclusions over marginal xGOT differences.
  • Each data provider uses a different model; for the same match, xGOT will not be identical across different websites, so comparisons must rely on the same source.

4Progressive Passes: The Most Easily Overlooked Definition

Progressive passes attempt to capture "meaningfully moving the ball toward the opponent's goal," rather than sideways or backward passes. The complication is that definitions differ across data providers:

Data ProviderOfficial Definition
OptaCompleted passes that move the ball at least 25% closer to the goal within the attacking two-thirds of the pitch (Source 1)
WyscoutForward passes that satisfy: at least 30 meters closer in one's own half; at least 15 meters across halves; at least 10 meters in the opponent's half; considered successful only if the next touch is by a teammate (Source 5)
FBrefCompleted passes that move the ball toward the opponent's goal line at least 10 yards (approx. 9.1 meters) from its furthest point in the last 6 passes, or any completed pass into the penalty area; excludes passes originating from the defending 40% of the pitch (Source 6)

(Opta and Wyscout measure closing "distance to the opponent's goal," with Opta's original phrasing being "at least 25% closer to the goal." FBref measures distance progressed toward the opponent's goal line.)

The exact same pass might count under Opta but not under Wyscout or FBref, and vice versa. Therefore:

  • Do not compare progressive pass numbers from Website A alongside those from Website B.
  • Comparisons of "who progresses more" can only be made within the same platform.

In addition, Opta separately tracks "progressive carries," defined as carrying the ball forward more than 5 meters (Source 1), so a player breaking out from the defensive third into midfield will still register progression, just not within the passing count.

How to Read It Meaningfully

In my view, progressive passes are best suited to answering "how this team / this player moves the ball forward," rather than "who is better":

  • High progressive passes for defenders and defensive midfielders: The team initiates attacks via buildup from the back, common in possession-oriented sides.
  • Many progressive passes but very low xA: The ball is frequently moved forward, but not into dangerous areas, likely consisting mostly of safe long balls.
  • Few progressive passes but high xA: The player passes less often but decisively, typical of final-ball attacking midfielders.

It is a descriptive measure of volume, not a quality rating, and no specific number constitutes a "passing grade." Like possession, it reveals match style, but should not be used directly to judge who played well.

5How to Look at the Three Metrics Together

What You Want to KnowLook at FirstPair With
Whether a team's chance quality is goodxGxPts (compared with points)
Whether a player has genuine creativityxAAssist tally (compared with luck), set-piece share
Whether a striker is clinical or luckyxG, xGOT, and goals side by sideSample size of matches
How a goalkeeper is performingFaced xGOT − goals concededGoals conceded, save percentage
How a team progressesProgressive passes, progressive carries (same platform)PPDA, possession

A simple example: when seeing an attacking midfielder with "9 assists in a season," first ask three things: (1) What is their xA? (2) Were the assists from open play or set pieces? (3) How was their teammates' xG conversion? You do not need to understand everything; simply developing the habit that "assists are the outcome, xA is the process" helps avoid most misjudgments.

6FAQ

What is the difference between xA and assists?

Assists are outcomes that require a teammate to score; xA measures the intrinsic value of the pass itself, regardless of whether the teammate scores. Over the long run, the two should be close; short-term discrepancies are often luck.

Is xGOT always higher than xG?

Not necessarily. Looking at a single on-target shot, one struck into the top corner will have a higher xGOT than the original xG; a shot hit straight into the keeper's arms may have an xGOT lower than the xG.

Nor can one simply say which is higher across a full season. A player's full-season xG sums "all shots," including off-target and blocked efforts; full-season xGOT only totals "on-target" shots, with off-target and blocked shots having an xGOT of 0 (Source 3). Therefore, full-season xGOT exceeding xG typically indicates that the player's on-target placement is better than average, a difference Opta terms shooting goals added (Source 3); for players who frequently shoot off-target or get blocked, full-season xGOT may actually be lower than xG. The above deduction is my own summary, and models vary across providers.

Why do figures differ across different websites?

Models, data feeds, and definitions all differ. When comparing players or teams, always use the same source.

Can progressive passes be used to compare defenders?

Only between defenders on the same platform, in the same season, and playing comparable roles. Teams with different tactical setups (such as long-ball-oriented versus short-passing buildup) naturally produce different figures, which does not indicate whose defenders are superior.

Can these stats predict match outcomes?

They describe events that have already occurred. To predict the next match, one must also account for opponents, injuries/suspensions, and home/away factors; if you want to try calculating win probabilities yourself, you can start with the Poisson Score Model.

7Sources

Last updated: 2026-10-18

Illustrated portrait of Old Parker

About the author

Old Parker

Football columnist

Editorial pen name

Old Parker covers the Premier League, Champions League, Europe's major leagues and international football. He reads matches through schedule load, home-field context and recent form, then uses the numbers to explain formats, qualification routes and viewing times for readers in Taiwan.

View profile and articles ›

‹ Back to guides · Glossary