A winger drops a shoulder, beats his man, and slides a pass to the eight. The eight takes a touch and threads it to the striker, who scores. The striker gets the goal; the eight gets the assist; the winger, whose dribble pulled the whole defence out of shape, gets nothing. Shot-creating and goal-creating actions exist to fix exactly that kind of accounting, by paying not just the player who made the assist, but the player who made the pass before it.

What SCA and GCA actually count

Shot-creating actions (SCA) and goal-creating actions (GCA) are a pair of metrics popularised through the public-analytics ecosystem: you will most often meet them on FBref, built on StatsBomb’s detailed event data. The idea is disarmingly literal. Take any shot. Now look at the two offensive actions immediately before it by the attacking team. Each of those two actions earns its player one shot-creating action. If the shot becomes a goal, the same two actions instead earn a goal-creating action.

The key word is two. Conventional credit stops at the assist: the single pass that set up the shot. SCA and GCA reach one step further back, to the action before the assist. That second action is where ball-carriers, line-breaking passers, and the players who win free-kicks finally show up in the box score.

The qualifying actions are a small, fixed menu. An action counts toward SCA or GCA if it is one of the following, and it is the last or second-last team action before the shot:

  • A live-ball pass: a pass in open play that helps create the shot.
  • A dead-ball pass: a pass from a set-piece such as a free-kick or corner.
  • A dribble (take-on): beating a defender to carry the move forward.
  • A drawn foul: being fouled in a way that leads to the shot, including winning the free-kick that is then delivered.
  • A defensive action: a tackle or interception that wins the ball back and begins the move.
  • The shot itself in the rebound case: a shot whose rebound or block leads directly to another shot.

A move, broken down: Messi, Gvardiol and Álvarez

Numbers make this concrete, so here is a real move: Argentina’s third goal against Croatia in the 2022 World Cup semi-final, in the 69th minute, taken from the StatsBomb event file (match 3869519 in StatsBomb Open Data (StatsBomb Public Data User Agreement)). Nahuel Molina took a throw-in, and Julián Álvarez passed to Lionel Messi. Messi beat Joško Gvardiol with a dribble just inside Croatia’s half, carried the ball 47.6 yards upfield into the penalty area, beat Gvardiol again near the goal line and cut the ball back to Álvarez, whose shot, worth 0.33 xG, went in.

Argentina’s third goal against Croatia, 2022 World Cup semi-final: every pass, carry, dribble and shot in the possession, and which of them earn shot- and goal-creating actions. Ball receipts left out. Names as StatsBomb records them. Data: StatsBomb Open Data (StatsBomb Public Data User Agreement).
#PlayerActionCredit
1Nahuel Molina LuceroThrow-in (dead-ball pass)None: qualifies by type, but too far back
2Julián ÁlvarezPassNone: qualifies by type, but too far back
3Lionel Andrés Messi CuccittiniCarry, 3.2 yards upfieldNone: carries never count
4Lionel Andrés Messi CuccittiniDribble past Joško Gvardiol, completedNone: qualifies by type, but too far back
5Lionel Andrés Messi CuccittiniCarry, 47.6 yards upfieldNone: carries never count
6Lionel Andrés Messi CuccittiniDribble past Joško Gvardiol, completed1 SCA, 1 GCA (second action before the shot)
7Lionel Andrés Messi CuccittiniCarry, 0.8 yards upfieldNone: carries never count
8Lionel Andrés Messi CuccittiniPass to Álvarez (the assist)1 SCA, 1 GCA (last action before the shot)
9Julián ÁlvarezShot, 0.33 xG, goalThe shot being credited
How the credit lands

The two actions directly before the shot were Messi’s pass (the assist) and, before it, his second dribble past Gvardiol. Both earn one SCA, and because the shot was scored, both earn one GCA: Messi takes 2 SCA and 2 GCA from a single move.

The 47.6-yard carry that took him from Croatia’s half into the penalty area earns nothing, because a carry is not on the menu; the first dribble, Álvarez’s pass and Molina’s throw-in are on it but sit too far back. SCA and GCA only ever look two deep.

That move also exposes the metric’s defining limit, which we return to below: it credits Messi twice for one move and gives nothing to its longest carry, purely because of where each action fell in the count and what type of action it was.

Shot-creating actions at the 2022 World Cup, by type

We applied the same rule to every shot at the 2022 World Cup: 1,453 shots in the 64 matches, penalties included and shoot-out kicks left out. For each one, our analysis script walks back through the match events from the shot and keeps the shooting team’s last two qualifying actions, stopping if the opponent had the ball in between. On StatsBomb’s event types the menu above becomes: a completed pass (dead-ball if it was a corner, free kick, throw-in, goal kick or kick-off, live-ball otherwise), a completed dribble, a foul won, a won tackle, an interception or a ball recovery, and an earlier shot. Carries and ball receipts never count. 1,395 shots had two qualifying actions behind them, 57 had one and one had none.

Shot-creating actions (SCA) and goal-creating actions (GCA) at the 2022 World Cup, by type of action: 1,453 shots, shoot-outs left out, our counting rule applied to StatsBomb’s events. Data: StatsBomb Open Data (StatsBomb Public Data User Agreement).
Action typeSCAShare of SCAGCA
Live-ball pass1,76662.0%210
Defensive action47816.8%55
Dead-ball pass2147.5%11
Shot (rebound case)1354.7%16
Drawn foul1344.7%27
Take-on (completed dribble)1204.2%13
All types2,847100%332

Live-ball passes are 62.0% of all shot-creating actions, which is why SCA leaderboards look like lists of passers. The second-largest group is the one the assist column never sees at all: tackles, interceptions and recoveries, 16.8% of SCA and 55 GCA. Take-ons, the action Messi used twice in the move above, are the rarest type, 4.2%. Lionel Messi and Kylian Mbappé tied for the most SCA at the tournament, 48 each; Mbappé had the most GCA, 11, and six players had 5, Messi among them. Those are totals, not rates, and both played seven matches.

Our counts follow the definitions on this page, applied to StatsBomb’s events; FBref’s published SCA and GCA are built from another provider’s data and will not match them exactly. Every figure in this section and the move above is in /data/stats/sca-gca-explained.json, written by our analysis script.

Why this widens the credit beyond the assist

The reason analysts reach for SCA is that the assist is a famously blunt instrument. It rewards one pass and ignores the action that made that pass possible: the carry that broke the lines, the dribble that committed a defender, the dead-ball delivery that fell to the assister. Across a season, two very different players can have near-identical assist totals while doing completely different jobs: one finishing moves with the final ball, the other repeatedly supplying the player who supplies the finish.

SCA pulls those two roles apart. A deep creator or a dribbling wide player who consistently makes the penultimate action will accumulate a high SCA total even with a modest assist count, because they are involved in the build-up of chances whether or not they make the final pass. It is the difference between asking “who set up the shot?” and “who were the last two players to move the ball toward this shot?” The second question has two answers, and that is the whole point.

Reading it per 90, and its place among creative metrics

Raw SCA and GCA totals are volume figures, so they reward minutes and they reward playing in a team that takes lots of shots. The standard fix is the same as for most event metrics: divide by minutes and quote them per 90. SCA per 90 tells you how often a player is one of the last two contributors to a shot in a typical match; GCA per 90 does the same for goals, but because goals are far rarer than shots, GCA per 90 is a small and noisy number over short samples: closer to a sanity check on SCA than a standalone leaderboard.

It helps to place SCA and GCA next to the creative metrics they complement rather than replace. Expected assists (xA) values the quality of the chance a player creates, the xG of the resulting shot, but only for the single creating pass. SCA values the act of creation for the last two players, but treats every shot as one unit regardless of how good the chance was. And xGChain and xGBuildup go to the opposite extreme, crediting every player in a possession with the full xG of the shot it produced, no matter how many actions back they were.

2
Actions counted, per shot
SCA
If it ends in a shot
GCA
If it ends in a goal

Read together, the three answer different questions. xA asks how dangerous your creation was; xGChain asks whether you were involved at all; SCA and GCA sit in between, asking whether you were one of the final two hands on the ball. A scout looking at a wide player would want all three: high SCA tells you they are repeatedly near the end of moves, xA tells you whether the chances they help make are any good, and xGChain tells you whether they matter to the build-up beyond those last two touches.

The caveats worth keeping

SCA and GCA are useful precisely because they are simple, and most of their problems are the flip side of that simplicity. First, the hard two-action cut-off is arbitrary: it credits the assist and the pass before it, but a brilliant third-action carry that set the whole move in motion gets nothing, while a trivial square pass that happens to fall in the last-two window gets full credit. The metric measures position in the sequence, not value.

Second, both are volume- and team-dependent. A player in a side that takes twenty shots a game has far more chances to register SCA than the same player in a side that takes eight, so cross-team comparisons need care and per-90 framing at a minimum. Third, GCA inherits all the noise of goals themselves: it depends on finishing, which the creating player does not control, so a low GCA next to a high SCA usually says more about teammates’ conversion than about the creator.

The honest way to use them, as with most of these numbers, is as a lens rather than a verdict. SCA and GCA do one valuable thing well: they drag the penultimate action into the light, and with it the players who quietly make chances without ever making the assist.

Sources, notes & further reading

Expected assists and expected threat

This section was first published on 14 May 2026 as a separate article.

Here is the problem with the assist as a statistic: it credits the wrong skill to the wrong player. Thread a defence-splitting pass to a striker who is clean through, watch him blaze it over the bar, and you get nothing on the scoresheet. Roll a lazy square ball to a teammate who taps into an empty net from six yards, and you get an assist: officially the identical creative contribution. An assist isn't really a measure of creativity at all; it is an accounting of other people's finishing, filed under your name. Expected assists (xA) and expected threat (xT) are the two metrics that try to fix that, and they fix it in different ways.

What expected assists actually measure

xA attaches the xG of the resulting shot to the player who created it, rather than to whether the ball happened to go in. The mechanic is almost insultingly simple, and that simplicity is the point: for every pass a player makes that leads directly to a shot, they bank the shot's xG value as credit. Sum those across any time period and you have their xA.

That means a player who consistently finds teammates in the penalty area with good looks at goal will accumulate high xA whether or not those teammates convert. The finisher's luck, or skill, is stripped out. A through-ball that sets up a 0.4 xG shot from the penalty spot counts as 0.4 xA, always, regardless of what happens next.

The contrast with assists is pointed. The most common ways raw assists inflate the wrong player: tap-ins where the "assist" was a routine corner; set-piece deliveries where the scorer did all the work with their head; deflected crosses that bobbled in off a defender. And the ways assists undercount real creators: the player who generated twelve 0.4 xG shots whose strikers missed every one, or the playmaker who was dispossessed the moment before a teammate scored and so appears on no record at all.

The 2022 World Cup xA leaders

The 2022 World Cup across 64 matches provides a clean sample with every chance built from StatsBomb's detailed event data. The xA leader was Antoine Griezmann, and he is probably not the first name that comes to mind when you think "World Cup creative force." He ended up with 2.54 xA, edging Lionel Messi at 2.45. Both numbers capture something the raw assist counts buried entirely.

Expected assists (xA) leaders, 2022 FIFA World Cup. xA = cumulative xG of shots the player directly created. Data: StatsBomb open data, retrieved June 2026.
Player xA
Antoine Griezmann2.54
Lionel Andrés Messi Cuccittini2.45
Leroy Sané1.40
Raphael Dias Belloli (Raphinha)1.40
Harry Kane1.36
Ousmane Dembélé1.29
Mateo Kovačić1.27
Bruno Miguel Borges Fernandes1.18
Hirving Rodrigo Lozano Bahena1.15
Mislav Oršić1.11

Griezmann leading that list is not a surprise to anyone who watched France closely. He spent much of the tournament operating in the half-spaces as a second striker, dropping deep to collect and then picking out runners. Most of that creation never made the assist column because French strikers spurned the chances. xA saw it anyway.

Bar chart of the ten players with the most expected assists at the 2022 World Cup, led by Antoine Griezmann.
Expected assists (xA) leaders at the 2022 World Cup: the total xG of the shots each player teed up. Data: StatsBomb open data, retrieved June 2026.
Pass map of Antoine Griezmann's chances created at the 2022 World Cup, with each assisting pass drawn as an arrow and the resulting shot sized by xG.
Antoine Griezmann's chance creation at the 2022 World Cup: every pass that led to a shot, with the shot location sized by xG. Data: StatsBomb open data, retrieved June 2026.

The limit xA runs into

xA is a significant improvement on raw assists, but it inherits one structural flaw: it only credits the player if their action directly preceded a shot. That last pass gets everything; the one before it gets nothing. Think about the midfielder who plays a quick turn-and-release out of pressure to an overlapping fullback, who then delivers the cross: the midfielder's contribution vanishes from the xA record. And an entire category of creative action is invisible to xA: carrying the ball through midfield, switching the play to open up the width, drawing defenders to create pockets for others. A box-to-box runner who breaks defensive lines with carries every other game and never touches the ball at the point of the shot simply doesn't exist in xA's world.

That is the gap expected threat was designed to fill.

Expected threat: valuing every action on the pitch

Expected threat (xT) is a possession-value model built around a simple question: given that a player has the ball in zone X, what is the probability that the team scores within the next few actions? The model assigns each point on the pitch a number between 0 and 1 that represents that probability. Any pass or carry that moves the ball from a lower-value zone to a higher-value zone gets credited with the difference. Any action that moves the ball backward, or gives it away, incurs a cost.

The intuition lands immediately. The areas near the opposing goal are worth a lot; deep in your own half, almost nothing. Nobody is going to argue with that. What xT makes rigorous is every gradient in between: the half-space outside the penalty area is more valuable than the same distance from goal directly down the wing, because the angle to shoot is better and defenders are split. The model learns those gradients from data, not from anyone's opinion.

How the xT grid is built

Divide the pitch into, say, 192 cells (16 columns × 12 rows). For each cell, estimate two probabilities from historical data: pshoot (how often a team with the ball here takes a shot within the next three actions) and pgoal | shot (how often those shots become goals). The cell's xT value is pshoot × pgoal | shot, plus a weighted sum over every zone the ball could move to next (a pass or carry), capturing the downstream value. Solve the system of equations and you get a value map. A pass from deep midfield to the edge of the penalty box might raise the ball from an xT of 0.04 to 0.13, crediting the passer with 0.09 xT.

A worked example

Suppose a central midfielder on the halfway line (xT roughly 0.02) plays a precise diagonal to a winger at the corner of the penalty area, where the xT value is approximately 0.10. That single pass earns the midfielder 0.08 xT. The winger then cuts inside to the penalty spot, a carry from xT 0.10 to maybe 0.20, earning 0.10 xT. The winger shoots (at that point, the shot is handled by the xG model rather than xT) and misses. Under xA, neither player gets anything. Under xT, the midfielder's diagonal and the winger's carry are both credited, capturing the genuine value they added to the attacking move. The miss is the striker's problem, not theirs.

That is the philosophical difference. xT is a model of progression; xA is a model of shot creation. They measure different things, and the most complete picture of a creator uses both.

Reading xA and xT together

A player who leads in xA almost certainly creates high-quality shots directly. A player who leads in xT might be driving dangerous ball progressions that end in shots for others: a deep-lying playmaker or an advancing fullback whose involvement is always three passes before the chance. High xT with modest xA is the signature of a player who shifts the structure of an attack rather than making the final key pass. High xA with modest xT could be an efficient final-third operator who contributes less to the build-up.

Neither stat is a verdict on a player's total value, and treating them that way is a mistake. xT can be gamed somewhat: if a player only ever receives the ball in high-value zones, their carry xT looks impressive for fairly modest effort. And xA, like xG, is noisy in small samples: a player can rack up 1.5 xA in one game and nothing across the next four. Both metrics stabilise with larger samples, and both are most useful as a diagnostic rather than a single-number ranking.

The old assist column told you who touched the ball last before a goal. xA and xT together tell you who made dangerous things happen and at what stage, irrespective of what a keeper or striker did about it. That is a truer account of creativity.

Sources, notes & further reading

  • StatsBomb open data: match events including shot xG and key-pass links used to derive the xA figures here.
  • StatsBomb: documentation on their event model and possession-value research.
  • Karun Singh's blog: the original public description of the xT framework.
  • FBref: xA and progressive passing data for domestic leagues.
  • Understat: xA and xG breakdowns for top European leagues.

Published by SoccerAnalytics from public data and the sources named in this piece. How this site is made.