Do Most Goals Come From Three Passes or Fewer? What the Passing-Sequence Research Actually Shows

Most goals do come from moves of three passes or fewer, but that does not mean short moves are the better tactic. Most moves of any kind are short. The claim began with Charles Reep's notebooks and has been argued over for more than half a century. On RubiScore, as anywhere, the useful question is what each move produces.

This data-study traces the claim from its origin to its critiques and asks what passing-sequence data can and cannot show. Live possession, passing and shot figures for current matches can be checked at https://rubiscore.com, while the historical studies below come from published research.

Where Did the Three-Pass Claim Come From?

It came from one of football's first systematic analysts. Charles Reep was an RAF accountant who rose to the rank of Wing Commander and spent decades notating matches by hand. His 1950s analyses in the News Chronicle caught the attention of Brentford manager Jackie Gibbons, who employed him as a part-time adviser from February 1951, and later of Stan Cullis at Wolverhampton Wanderers.

The formal version came in 1968. Reep and the statistician Bernard Benjamin published "Skill and Chance in Association Football" in the Journal of the Royal Statistical Society, Series A. Their data covered English league football from 1953 to 1967, a total of 578 games: 101 games in a first set and 477 in a second.

The headline finding concerned the moves themselves. Only 5% of moves contained four or more passes, and only 1% contained six or more. Reproductions of their distribution show the most common outcome was a move that ended before a single pass was completed, at 39.4%, followed by one pass at 27.5% and two passes at 16.5%. Richard Pollard later noted that the counts closely followed a negative binomial distribution, a sign that chance plays a large part in how long a move lasts.

From this Reep concluded that most goals came from moves of three passes or fewer, and that possession football was therefore undesirable. The idea travelled. Charles Hughes, director of coaching at the Football Association, built on it in his book The Winning Formula, with its "Positions of Maximum Opportunity" (POMO), areas where he argued goals would follow if the ball arrived there often enough. Reep's ideas also shaped the style of the Norway national team under Egil Olsen in the 1990s.

What Does Passing-Sequence Data Actually Capture?

A passing sequence, or move, is a run of consecutive completed passes by one team before something ends it. The count is simple enough to do with a pencil, and that simplicity limits what it can say.

A pass count records:

• Length. How many completed passes came before the move ended.
• Outcome. Whether the move ended in a shot, a goal or a loss of the ball.

It leaves out most of what a modern analyst would want to know:

• Where the move started. A regain near the opponent's box and a goal kick are both "zero-pass" starts.
• How the ball travelled. Carries and dribbles move the ball without adding a pass.
• Time and speed. Four quick passes and four slow ones look identical.
• The defence faced. Whether the opponent was organised or caught out of shape.

Every later argument about Reep's numbers turns on these missing fields.

Why Is "Most Goals Came From Short Moves" Misleading?

Because it reports where goals were found without saying how often each kind of move happened. This is the base-rate problem, and the writer Jonathan Wilson set it out in his 2009 history of tactics, Inverting the Pyramid.

Wilson's reading of Reep's figures was that 91.5% of moves had three passes or fewer. If short and long moves were equally productive, roughly 91.5% of goals should also have come from short moves. Reep's analysis found that fewer than 80% did. On Reep's own numbers, then, moves of four or more passes produced more than their share of goals. Wilson concluded that, move for move, short sequences were less effective.

A hypothetical example makes the arithmetic clear. Imagine a season of 10,000 moves in which 9,000 are short and 1,000 are long. Suppose short moves produce 180 goals and long moves produce 40. Short moves "account for most goals", 180 out of 220, yet each short move scores at a rate of 2 in 1,000, while each long move scores at 4 in 1,000. The figures are invented to show the logic.

What Did Hughes and Franks Find When They Normalised the Data?

The best-known attempt to correct for base rates came from Mike Hughes and Ian Franks. Their 2005 paper, "Analysis of passing sequences, shots and goals in soccer" in the Journal of Sports Sciences, examined passing sequences at two FIFA World Cup finals tournaments, 1990 and 1994.

Their key step was normalisation: they divided the goals from sequences of each length by the number of sequences of that length. Their reported findings were:

• More goals per sequence from longer moves. Once frequency was accounted for, longer passing sequences produced more goals than shorter ones.
• More shots per possession. Longer sequences generated significantly more shots per possession.
• Better conversion for direct play. The ratio of goals to shots was better for "direct play" than for "possession play".
• A split by team quality. At the 1990 World Cup, longer sequences produced more goals per possession for successful teams, while neither approach clearly helped unsuccessful teams.

The authors concluded that Reep and Benjamin's work, although a landmark in match analysis, gave only a partial understanding of how goals are scored.

Why Is the Normalised Answer Not the Final Word Either?

Richard Pollard, who had worked with Reep, challenged that reading in a 2019 letter in The Open Sports Sciences Journal, titled "Invalid Interpretation of Passing Sequence Data to Assess Team Performance in Football". First, he says the way Hughes and Franks defined a sequence ignored several ways a move can end, including shots, tackles, fouls and the ball going out of play.

Second, most sequences are short whatever style a team plays. According to Pollard, fewer than 20% of sequences reached five passes, so labelling every move of up to four passes as "direct play" is, in his word, meaningless. Many of those short moves are possession play that failed early, not direct play that succeeded or failed on its own terms.

Pollard's wider point is that passing-sequence length cannot settle whether direct or possession play is more effective, in either direction. That cuts against Reep's original reading as much as against the rebuttal. Pollard instead pointed to outcome-based measures. Reep's own work found that from midfield, a long forward pass into the attacking third was on average twice as likely to produce a shot as short passing. Pollard and Reep later measured "yield", the goals expected to be scored minus those conceded per 1,000 actions. A long forward pass approach had a yield of 23.1, against 11.1 for approaches built on short passes only, and a corner had a yield of 23.

The method matters more than the old figures: it values an action by what tends to follow it, the same principle modern possession-value models apply to far richer data.

What Else Confounds the Question?

Several confounders sit between a style of play and a goal.

• The state of the defence. Tenga, Holme, Ronglan and Bahr (2010) studied 163 of 182 matches in the 2004 Norwegian top flight. They compared 203 possessions that led to goals with 1,688 random possessions as controls. Counterattacks made up 52% of the goal-scoring possessions but only 41% of the controls. Against an imbalanced defence, counterattacks were more effective than elaborate attacks (odds ratio 1.64). The defence faced, not the pass count, carried the signal.
• Team quality. Collet (2013) analysed five European leagues plus UEFA and FIFA tournaments from 2007 to 2010. Possession and passing predicted aggregated team success in domestic leagues, but at the level of individual matches, after accounting for team quality and home advantage, both were poor predictors. Strong teams have the ball more, so any style they favour inherits their results.
• Game state. Teams chasing a game go long more often, and teams protecting a lead keep the ball. Pooling all minutes mixes these situations together.
• Era and level. Reep's data came from English football between 1953 and 1967, and Hughes and Franks used World Cup finals. Neither describes a modern league.

How Should Passing-Sequence Numbers Be Read Today?

Pass counts still have a use, as a description of a team's style rather than a verdict on it. Match statistics on platforms such as RubiScore, including possession, total passes and shots, show how a team played in a given match. Turning that into a judgement about effectiveness takes a few checks:

• Compare rates, not shares. Ask how often each type of move produces a shot or a goal, not what share of goals it accounts for.
• Separate starting points. Moves that begin with a high regain, a goal kick or a set piece answer different questions.
• Account for opponent and game state. The same style meets very different defences at 0-0 and at 2-0 down.
• Prefer outcome-based measures. Shot quality and possession-value models judge actions by what they lead to.

Verdict: What Does the Evidence Support?

The narrow claim is true. In the data Reep collected, most goals came from moves of three passes or fewer. The tactical conclusion drawn from it does not follow, because short moves were simply the most common kind. On Reep's own figures, as Wilson showed, longer moves were more productive per move.

The rebuttal has limits too. Pollard argued that pass counts cannot separate the two styles at all, because many short sequences are failed possession play. The research does point clearly at what matters more than pass count: the state of the defence when the ball is won, the quality of the team, and where on the pitch the move begins.

The honest answer, then, is that the number of passes is a poor guide to whether a style of play works. It describes how a goal was scored, and it cannot show why.