For most of football's history the goalkeeper had one job, and we measured it with one number: did the shot go in or not? That keeper still exists. But sitting on top of him is a second job (first passer, last defender, the player who turns a goal kick into the first phase of an attack), and that job leaves a trail in the data. This is how analysts read it.

From the line to the build-up

The modern goalkeeper's role has expanded in two directions at once. Forward, into the team's possession: the keeper is now the deepest outfield player when his side has the ball, an extra body who lets a back four become a back five in build-up and break a high press by playing through it rather than over it. And outward, away from goal: when the defensive line pushes up to compress the pitch, someone has to guard the grass behind it, and that someone is the keeper, racing off his line to head clear or smother a through-ball before it becomes a chance.

Neither of these is captured by the thing we traditionally graded keepers on. Shot-stopping is its own discipline, and the right tool for it is post-shot xG, which compares the chances a keeper faced with the goals he actually conceded. We cover that in full in post-shot xG and goalkeeper metrics. This article is about everything else: the passing, the sweeping, and the risk that comes bundled with both. Those, too, are now measurable.

Measuring distribution

Start with the ball at the keeper's feet. The first thing event data records is what he does with it, and a handful of metrics turn that into a profile.

  • Pass completion %. The blunt headline number, but it needs context: a keeper who only ever rolls it to the nearest centre-back will post a gaudy completion rate while contributing nothing to progression. Completion is only meaningful read alongside where and how far the passes go.
  • Long-vs-short distribution. What share of a keeper's passes travel a long distance (a common cutoff is roughly 35 yards or more) versus stay short? A keeper at 20% long is being used as a build-up outlet; one at 70% long is mostly clearing his lines or aiming for a target man.
  • Launch %. A close relative: the proportion of goal kicks and open-play distributions that are simply hit long downfield. High launch percentage is the statistical fingerprint of a team that does not play out from the back, whether by design or by necessity.
  • Progressive passing. The richest view borrows the framework from outfield play (see progressive passes and carries) and asks how much territory the keeper's distribution actually gains, and how much of it survives the opponent's press.

The crucial qualifier on all of these is under pressure. Anyone can pass to a free centre-back in an empty six-yard box. The keeper who matters in a possession side is the one whose completion barely dips when two forwards are bearing down on him, because that composure is what lets his team commit numbers forward and trust the ball will come out clean.

It helps to keep two questions apart: what does the keeper choose to do, and how well does he do it. Launch % mostly answers the first, and that choice is a tactical decision made above the goalkeeper's head. A 70% launch rate is a failure in a possession side and exactly the plan in a direct one (same number, opposite verdict), so establish the style before grading the skill, and compare keepers within similar systems. Watch the scoreline too: a team protecting a lead launches more to kill the game, so a full-match launch % can overstate how direct a keeper naturally is, the same game-state effect that warps possession and pressing numbers.

Measuring the sweeping

The second half of the job, the ground covered outside the box, is harder to see but not invisible. Two ideas do most of the work.

Defensive actions outside the penalty area. Every time a keeper makes a clearance, an interception, or a tackle beyond his box, the event is logged with a location. Count those, and you have a direct measure of how often he acts as an auxiliary defender rather than waiting on his line.

Average distance of defensive actions from goal. This is the single best public proxy for "sweeping". Take all of a keeper's defensive actions, measure how far from his own goal each one happened, and average them. A high number means a keeper who habitually patrols the space in front of his box; a low number means one who stays home. It is only a proxy (it says nothing about whether the sweeping was necessary or merely flashy), but as a one-glance descriptor of playing style it is remarkably informative.

Twenty-eight World Cup keepers, measured

To put real names on those metrics, we measured every goalkeeper who played at least 270 minutes (three full matches) at the 2022 World Cup in StatsBomb Open Data (StatsBomb Public Data User Agreement): 28 keepers. A long pass is one of 40 yards or more, goal kicks included. A sweeper action is StatsBomb’s “Keeper Sweeper” event, the keeper leaving his line to claim or clear a ball played in behind; distance is measured from the centre of his own goal, in yards (StatsBomb’s pitch is 120 by 80). The table shows the five keepers who went long least and the five who went long most.

2022 World Cup goalkeepers with 270 or more minutes: the five lowest and five highest shares of long passes (40+ yards) among the 28. Sweeper actions = StatsBomb Keeper Sweeper events (outside the penalty area in brackets); distance = their average distance from the centre of the keeper’s own goal. Data: StatsBomb Open Data (StatsBomb Public Data User Agreement), computed by our analysis script.
KeeperTeamMinutesPasses per 90Long passesCompletedSweeper actionsDistance (yd)
Unai SimónSpain41434.48.2%93.7%3 (3)30.9
AlissonBrazil39521.416.0%85.1%1 (1)31.7
Yann SommerSwitzerland28636.817.9%91.5%2 (0)13.2
Manuel NeuerGermany29437.720.3%90.2%4 (3)25.2
Thibaut CourtoisBelgium28438.623.0%89.3%1 (1)27.9
Hugo LlorisFrance61426.759.3%61.0%4 (1)16.1
Keylor NavasCosta Rica29430.959.4%54.5%6 (1)13.3
Aymen DahmenTunisia29723.462.3%62.3%2 (0)13.4
Mohammed Al OwaisSaudi Arabia29929.565.3%54.1%7 (4)25.2
Wojciech SzczęsnyPoland39033.767.8%60.3%3 (0)14.0

The libero: Unai Simón. Spain’s keeper played 34.4 passes per 90, only 8.2% of them long, and completed 93.7%, including 23 of the 25 he played under pressure. All three of his sweeper actions came outside the box, on average 30.9 yards from goal. Germany’s Manuel Neuer, the keeper the term is most often attached to, fits the same profile: 20.3% long, 90.2% completed, and three of his four sweeper actions outside the box, at an average of 25.2 yards.

The long kicker: Wojciech Szczęsny. Poland’s keeper sent 67.8% of his passes 40 yards or more, the highest share of the 28, completed 60.3%, and averaged 49.5 yards a pass. His three sweeper actions all stayed inside the box, an average of 14.0 yards from goal.

But the two halves of the job are separate. Saudi Arabia’s Mohammed Al Owais went long with 65.3% of his passes and still made more sweeper actions per 90 than any other keeper, 2.11 (seven in three matches, four of them outside the box, at 25.2 yards). Brazil’s Alisson went long with only 16.0% of his passes and made one sweeper action in 395 minutes. Across the 28, the long-pass share and sweeper actions per 90 correlate at r = +0.33: if anything, the keepers who kicked long swept slightly more, not less. The neat “libero versus shot-stopper” split is a starting point, not a law.

Two cautions. First, completion is mostly a restatement of how far a keeper kicks: across the 28, long-pass share and completion correlate at r = -0.94, which is why completion has to be read alongside length, as the list above says. Second, sweeper actions are rare. The 28 keepers made 92 between them, 28 outside the box, between 0.19 and 2.11 per 90, so a keeper’s average distance rests on one to seven events and describes three to seven matches, not a settled skill. Spain’s and Poland’s systems chose much of what their keepers did; the numbers show the style first and the skill only second.

The trade-off: possession value versus risk

Here is the catch that makes the sweeper-keeper an analytical problem rather than a cheerleading exercise. Every short pass played out from the back under pressure carries a small probability of disaster: a misplaced ball, a heavy touch, a pressing forward poking it loose six yards from goal. When that happens the resulting shot is worth a huge slice of xG, because almost nothing converts as reliably as a chance handed to an attacker in front of an open net.

So the keeper's distribution is genuinely a risk-reward gamble. Playing out adds value (it beats the press, retains the ball, and starts attacks from deep), but it occasionally concedes the highest-quality chances in the game. The honest way to evaluate a build-up keeper, then, is not to celebrate completion percentage in isolation but to weigh the possession value created against the xG conceded from errors. A keeper who completes 92% of his passes but gifts three shots a season from his own box may be a worse fit for a given team than a less stylish one who simply does not make that mistake.

This is also why the eye test and the data so often disagree about goalkeepers. A spectacular sweeping clearance is memorable; the slow accumulation of a few extra metres of progression per game is not. And a single calamitous error is seared into the highlight reel, while the dozens of presses it broke are forgotten. The metrics exist precisely to put those forgotten contributions back on the ledger.

Does playing out from the back pay? 4,181 keeper passes followed

The trade-off above can be put to the 2022 World Cup. We took every pass a goalkeeper played in the 64 matches, 4,181 of them by 41 keepers (970 goal kicks among them), split them at 40 yards as in the keeper table further up, and followed each one twice. Forward: did the keeper’s team keep going, get into the opponent’s half, and finish the possession with a shot? Backward: did the opponent take a shot within 20 seconds of the pass?

Every goalkeeper pass at the 2022 World Cup, all 64 matches (extra time in, shoot-outs out). Short = under 40 yards. “Reached the opponent’s half” and “Ended in a shot” use the rest of the keeper’s team’s possession after the pass; own xG is the non-penalty xG of those shots. Conceded = an opponent non-penalty shot within 20 seconds of the pass. Under pressure is StatsBomb’s own flag on the pass. Data: StatsBomb Open Data (StatsBomb Public Data User Agreement), 2022 World Cup event files, retrieved June 4, 2026; computed by our analysis script.
Keeper passesPassesCompletedReached the opponent’s halfEnded in a shotOwn xG per 100Followed by an opponent shotxG conceded per 100Goals conceded
Short, all2,48397.4%71.4%8.2%0.89370.164
Long (40+ yards), all1,69838.9%76.9%5.6%0.81220.194
Short goal kicks48299.4%69.9%8.7%1.0340.060
Long goal kicks48839.5%85.0%6.1%0.7550.181
Short, under pressure14586.9%43.4%5.5%0.4371.042
Short, not under pressure2,33898.1%73.1%8.3%0.92300.112

Short was not riskier overall. Per 100 passes, short ones were followed by 0.16 xG conceded within 20 seconds and long ones by 0.19; 4 goals followed short passes and 4 long ones. Going forward, short passes led to 0.89 xG per 100 against 0.81 for long ones. Goal kicks show the same picture more sharply: 482 short goal kicks were completed 99.4% of the time, led to 1.03 xG per 100 and were followed by 0 goals conceded, against 39.5%, 0.75 and 1 for the 488 long ones.

But much of that is who plays short. The teams that played short were not a random draw. Within teams, the edge disappears: of the 31 teams with 20 or more keeper passes of each kind, 15 created more xG per 100 after their short passes than after their long ones, and the rest did the opposite. On this evidence the method itself is close to neutral; the team behind it decides most of the return.

The risk sits under pressure. Only 145 of the 2,483 short passes (5.8%) were played with an opponent pressing the keeper. They were completed 86.9% of the time against 98.1%, reached the opponent’s half 43.4% of the time against 73.1%, and 7 of them (4.8%) were followed by an opponent shot within 20 seconds, against 30 of 2,338 (1.3%) unpressed. Per 100 passes that is 1.04 xG conceded against 0.11, about nine times as much, and the balance of own xG minus xG conceded turns from +0.81 to −0.61. These are the seven:

The 7 short goalkeeper passes under pressure at the 2022 World Cup that were followed by an opponent non-penalty shot within 20 seconds. Clock is StatsBomb’s match minute and second; xG is the total of the opponent’s shots in that window. Data: StatsBomb Open Data (StatsBomb Public Data User Agreement), 2022 World Cup event files, retrieved June 4, 2026; computed by our analysis script.
KeeperMatchClockPass completedOpponent xGResult
Unai SimónJapan v Spain47:41Yes0.03Goal
Vanja Milinković-SavićBrazil v Serbia45:32Yes0.42No goal
Diogo CostaSouth Korea v Portugal66:21Yes0.10No goal
Unai SimónSpain v Germany23:54No0.03No goal
Mathew RyanFrance v Australia31:08Yes0.78Goal
Hugo LlorisEngland v France70:15Yes0.01No goal
Dominik LivakovićArgentina v Croatia81:53No0.14No goal

Two of the seven ended in goals: Spain conceded against Japan within 20 seconds of Unai Simón’s pass at 47:41, which had found a team-mate, and Australia against France within 20 seconds of Mathew Ryan’s at 31:08. In five of the seven the pass itself was completed and the shot came later in the move. That is the honest version of the risk: rare, concentrated in the few passes played into a press, and not always the keeper’s fault, since a window like this one also counts shots his pass only started.

Method and limits. All 64 matches of the 2022 World Cup from StatsBomb Open Data (StatsBomb Public Data User Agreement), event files retrieved June 4, 2026; extra time is included and shoot-outs are left out. A 20-second window catches shots that follow a keeper’s pass, not only shots it caused, and own xG counts the rest of the possession however long it ran, so the two halves are measured differently and only the comparisons between rows are like for like. Pressure is StatsBomb’s flag, not a count of pressing players. And seven bad outcomes in a tournament is too few to rank keepers by.

Reproduce it. Our analysis script computes every figure in this section; its output, including the within-team comparison, is in /data/stats/the-sweeper-keeper-quantified-risk.json.

Why the role exploded: rules and ideas

Two forces turned the sweeper-keeper from a quirk into a near-requirement. The first was a rule change. In June 2019 the laws were amended so that a goal kick is in play the moment it is taken and may be received by a team-mate inside the penalty area, rather than having to leave the box first. Overnight, the cleanest way to beat a high press (drop a centre-back into the box, let the keeper roll it short, and build from a numerical overload) became legal and routine. The change did not invent playing out from the back, but it removed the last bit of friction and accelerated its spread down the divisions.

The second force was tactical fashion. The high-pressing, possession-dominant football associated with the era's defining managers demands a keeper who can function as a tenth outfield player. Once the best teams in the world were winning with build-up keepers, the model propagated everywhere, and recruitment departments started screening goalkeepers on their feet as seriously as on their hands.

One final caveat haunts these numbers, as it haunts every per-90 statistic. A keeper at a dominant team passes in comfort; a keeper at a side camped on its own edge does everything under siege. Raw distribution and sweeping totals are shaped by the team in front of the keeper as much as by the keeper himself, which is why they travel better when possession-adjusted (see possession-adjusted stats) so that volume of opportunity does not masquerade as quality.

Reproduce it

Our analysis script computes every figure in the keeper section from StatsBomb’s event files for the 64 matches of the 2022 World Cup (the match list in StatsBomb Open Data), retrieved June 4, 2026, under the StatsBomb Public Data User Agreement. Minutes come from the event clock (starting line-ups and substitutions, stoppage time included; penalty shoot-outs left out); passes, pass lengths and outcomes, pressure and Keeper Sweeper events are StatsBomb’s own. All 28 keepers’ figures are in data/stats/the-sweeper-keeper-quantified.json.

Sources & further reading

  • StatsBomb Open Data: the 2022 World Cup event files behind the keeper section (StatsBomb Public Data User Agreement), with goalkeeper actions, locations, and pass detail to build distribution and sweeping profiles.
  • The keeper-pass follow-up (added November 11, 2026): the same StatsBomb Open Data event files (StatsBomb Public Data User Agreement), with possessions and the under-pressure flag as defined in StatsBomb’s event specification. Our results: data/stats/the-sweeper-keeper-quantified-risk.json.
  • StatsBomb: documentation of goalkeeper event definitions, including distribution and defensive-action types.
  • FBref: advanced goalkeeping tables (launch %, average pass length, defensive actions outside the box) for major competitions.
  • Understat: season-level data for the major European leagues.

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