A club with no tracking system, no camera contract and no analyst on the payroll can still measure how aggressively a team defends. It needs one thing: an event feed with a team label, an event type and a pitch coordinate. Everything else is arithmetic.
The measure built on those three fields is passes allowed per defensive action. It is cheap, it is reproducible, and it has been used to describe pressing for long enough that most people quoting it have stopped asking what it counts. That is the problem worth addressing, because there are three specific match situations in which the number says the opposite of what happened.
What follows is the calculation, the definition choices hidden inside it, the three failure modes, and two complementary numbers that a club without tracking can compute from the same feed on the same afternoon.
The metric on one page
- What it is. Opposition passes attempted in a defined area, divided by the defending team’s defensive actions in that same area.
- Direction. Lower is more aggressive. A team allowing eight passes per action is pressing harder than one allowing eighteen.
- Inputs. Event type, team, and a pitch coordinate along the length of the field. Nothing else.
- Defensive actions counted. Tackles, interceptions, challenges and fouls. Definitions vary by provider, and so does the number.
- The area. Conventionally everything outside the defending team’s own defensive portion of the pitch, usually the first forty per cent of its length.
- The rule to remember. It measures where the defending happened. It does not measure whether it worked.
What the number actually counts
The measure is a ratio, and both halves of it are counts of things that happened rather than judgements about quality. The numerator counts opposition passes in the zone. The denominator counts the defending team’s attempts to interrupt them in the same zone. Divide the first by the second and you get how many passes the opposition was allowed to make between interruptions.
That is a narrower claim than the one usually attached to it. It does not say the press was successful, because a defensive action includes a foul and a failed tackle. It does not say the press was coordinated, because a single player chasing the ball around produces the same events as an organised trap. And it does not say anything at all about what happened in the defending team’s own third, because that area is deliberately excluded.
Read strictly, the number answers one question well: over the whole match, how often did this team try to interrupt the opposition high up the pitch, relative to how much the opposition passed up there. Everything beyond that is interpretation, and most of the misreadings come from interpretation that the arithmetic does not support.
Where the PPDA pressing metric comes from, and what it inherits
The PPDA pressing metric was built to be computable from the event data that already existed, at a time when tracking data was expensive and rare. That origin explains both its usefulness and its limits: it inherits every property of an event feed, including the ones nobody likes.
The first inheritance is coding variability. Tackles, challenges and interceptions are the fields where two human coders watching the same match agree least, because the boundary between a tackle and a block, or between an interception and a loose ball collected, is a definition rather than an observation. A metric whose denominator is built almost entirely from those fields carries that noise straight through.
The second inheritance is definitional drift. Providers change their event taxonomy between seasons, and a change to how a challenge is recorded moves every historical value of the metric without moving anything on the pitch. Any comparison across providers or across a taxonomy change is unsafe, and the wider growth of commercial data operations has made multiple versions of the same match more common, not less.
Calculating it from an event feed, field by field
The computation takes a few lines of code or an afternoon in a spreadsheet. The steps below assume a feed with one row per event, a team identifier, an event type and coordinates normalised so that each team attacks from zero towards one hundred.
- Split the feed by possession team. You are computing one value for each side, and each value uses the other side’s passes.
- Normalise direction. Flip the coordinates for one half so that both halves describe attacking direction consistently. Skipping this is the most common error in a first attempt.
- Filter the numerator. Count opposition passes attempted with an origin coordinate above the zone threshold, conventionally forty on a hundred-point scale measured from the defending team’s own goal.
- Filter the denominator. Count the defending team’s tackles, interceptions, challenges and fouls with a coordinate in the same zone.
- Divide. Numerator over denominator. Report it to one decimal place and no further; the precision is not there.
- Report the denominator alongside it. A ratio built on nine defensive actions is not the same object as one built on forty, and printing only the ratio hides that.
The sixth step is the one that separates a usable analysis from a quotable number. Most of the disagreements about a value dissolve once both people can see how many events it was computed from.
The zone definition is the whole argument
The threshold is a convention, not a law, and moving it changes the story. Set it low and the metric absorbs midfield defending, which makes almost every team look aggressive. Set it high and it describes only the front line’s work, which produces small counts and violent match-to-match swings.
There is no universally correct answer, and there is a correct practice: pick a threshold, write it down, and never mix values computed under different ones. A club that compares its own number with a published figure without knowing the other party’s threshold is comparing two different metrics that happen to share a name.
The same applies to which events count as defensive actions. Some implementations include fouls, some exclude them, and some add ball recoveries. Every variant is defensible and none of them are interchangeable, which is why any internal analysis needs its own written glossary before it needs a second season of data.
Misread one: the team that never gets the chance to press
The first failure case is a team that dominates the ball. If the opposition has possession for a small share of the match and spends much of it in its own defensive third, the number of opposition passes inside the zone is small and the number of defensive actions is smaller still.
The ratio that comes out of that is unstable. Two or three extra interceptions in a match with a thin denominator swing the value substantially, and the swing has nothing to do with a change in pressing intent. Across a season these values average into something meaningful; across one match they are close to noise.
The mirror image is a team that defends deep by design. It records very few defensive actions in the zone because it is not there, and very few opposition passes in the zone because the opposition walks the ball forward unopposed and then plays it into the final third. The ratio can come out looking moderate, describing a team that never pressed at all. Neither case is a flaw in the arithmetic; both are cases where a ratio is being asked to describe a situation it cannot see.
Misread two: game state does most of the work
Pressing intensity is not a constant property of a team. It is a decision that changes with the score, and a full-match value averages across two different teams wearing the same shirts.
A side that leads from the twentieth minute will drop its defensive line, concede the ball in midfield and stop committing players forward. Its value for the remaining seventy minutes will be far less aggressive than for the first twenty, and the published match figure will sit somewhere between the two, describing neither.
The correction is straightforward and rarely applied: split the match by score state and report the value for each. Doing so usually reveals that two teams with identical match figures pressed in completely different circumstances, and it is the single largest improvement available to anyone using the metric seriously. A tactical account such as a breakdown of a knockout tie reads very differently once the numbers are cut this way. This is a textbook case of confounding: the score is driving both the pressing and the result being explained by it.
Misread three: an opponent who bypasses the zone entirely
The third failure comes from the other team’s style. A side that plays long from the goalkeeper, wins the second ball and attacks from there makes very few passes inside the zone. The passes it does make are long ones, and a long pass counts exactly the same as a short one in the numerator.
Against that opponent, a pressing team can produce an aggressive-looking value having barely pressed, because the small numerator did the work. It can equally produce a passive-looking one, because a long ball that clears the zone gives nobody anything to interrupt. Either way, the value is describing the opponent’s distribution policy rather than the defending team’s behaviour.
The tell is in the pass length distribution, which most feeds carry. A match where the opposition’s average pass length in its own half is unusually high is a match where this metric should be reported with a caveat or not at all. Teams built to play through a press, as national sides working on a possession-based tactical model generally are, produce the cleanest values against it; direct opponents produce the dirtiest.
Complement one: the height of the defensive action
The first cheap supplement uses data the club already has and answers a question the ratio cannot: where on the pitch does this team win the ball back.
Take every defensive action by the team, take its coordinate along the length of the pitch, and compute the average. That single number tells you the height of the defensive block far more directly than any ratio, and it does not depend on how much the opponent passed. Compute the same average for the actions that actually resulted in the team keeping the ball and the second number describes effectiveness rather than intent.
The pair is genuinely informative. A high average height with a low regain rate describes a press that is being played through. A low average height with a high regain rate describes a compact block doing exactly what it was asked to do. Neither statement is available from the ratio alone, and both cost nothing beyond a column of coordinates.
Complement two: time to the first defensive action after a turnover
The second supplement needs one extra field, a timestamp, and it measures the reaction rather than the position. After the team loses the ball, how many seconds pass before it makes its first attempt to win it back.
Compute it as a median across all turnovers in the match, and split it by the zone in which the ball was lost. A team with a genuine counter-pressing instruction produces a short median after losses in the attacking half and a longer one after losses deep. A team that retreats on every turnover produces a flat distribution.
The value of this number is that it is difficult to fake and easy to interpret. It is also robust to the three failure cases above, because it is conditioned on an event that definitely happened rather than on a ratio of two counts that may both be small. Clubs building an internal capability in performance data analysis get more from this measure in the first month than from any model.
Reading the three numbers together
| Measure | Fields required | What it answers | Where it breaks |
|---|---|---|---|
| Passes allowed per defensive action | Team, event type, coordinate | How often the team interrupted, relative to opposition passing | Small denominators; game state; direct opponents |
| Average height of defensive actions | Team, event type, coordinate | Where the block sits along the pitch | Says nothing about success |
| Height of successful regains | Adds an outcome flag | Where the team actually wins the ball | Outcome flags are coder judgements |
| Time to first defensive action | Adds a timestamp | How quickly the team reacts to losing the ball | Needs clean possession boundaries |
Used together, these four describe a defensive approach reasonably well without a single frame of tracking data. Used alone, the first of them supports about a third of the claims that are routinely made from it, and the other two thirds are where analysis gets a bad name. A general awareness of selection bias is more useful here than another metric.
Frequently Asked Questions
What counts as a low value?
It depends on the competition, the season and the provider, which is why quoting a universal threshold is a mistake. The usable approach is to compute the distribution for the league you are working in, from one provider, and read any team’s value as a position within that distribution rather than against a remembered figure from elsewhere.
Should fouls be included in the denominator?
Both choices are defensible. Including them captures the tactical foul as a legitimate act of pressing. Excluding them stops a team that fouls constantly from appearing to be an outstanding pressing side. Pick one, document it, and report the foul count separately so a reader can see which effect is present.
Can the metric be computed for a single half or a single spell?
Yes, and it should be, with one condition: report the denominator. Over a twenty-minute spell the counts can fall to single figures, at which point the ratio is an illustration rather than a measurement, and it should be presented as such.
Does it work in low-scoring, low-volume leagues?
It works less well, for the same reason most rate statistics do. Fewer events means a wider error band around every value, and a difference between two teams that would be meaningful in a high-volume league may be inside the noise in a low-volume one. Pooling several matches before drawing a conclusion is the practical answer.
What does a club need to start computing it?
A season of event data from one provider, a spreadsheet, and a written definition of the zone and the event types. No tracking system, no subscription to a modelling platform and no analyst with a statistics degree. The discipline of writing down the definitions is worth more than any of those.
Is it worth using at youth level?
Only if the data collection is consistent, which at youth level it rarely is. A coach hand-coding his own matches will produce a usable internal series if he applies the same definitions every week, and comparing that series with anything published elsewhere will not be meaningful. Internal trend, yes; external comparison, no.
The metric is a good tool used for the job it was designed for and a poor one used as a summary of a defensive philosophy. Report it with its denominator, split it by score, and put a second number beside it, and it stops making claims it cannot support.
