A recovery run gets logged only when it fails to work

The sprint that deters a through ball leaves no trace in the data, so the players who deter best look like they do least.

The run that stops the pass being played is never recorded.

Think about what event data actually captures. A tackle. An interception. A block. All of them require contact with the ball, which means all of them require the attacking move to have proceeded far enough to produce contact. A defender who sprints forty metres back, arrives in the passing line, and causes the ball carrier to choose a different, safer option, generates precisely nothing. No event. No entry. The database records that nothing happened, which is true, and misses that somebody made nothing happen.

This is a selection effect at the level of measurement itself, and it is stubborn, because you cannot fix it by collecting the same data more carefully. The event never existed to be collected.

The practical consequence is a systematic bias in how defensive contribution is valued. Players whose recovery work is fast enough to remove options look quiet. Players who arrive slightly later, and therefore have to make a visible tackle, accumulate a highlight reel and a statistical profile. If two full backs cover the same ground with the same intelligence, the marginally slower one has the better numbers.

Tracking data can address this and mostly is not used to. The information is there: you can measure how much of the pitch a defender denied, how the passing options available to the ball carrier shrank in the two seconds after a recovery run began, whether a threatening lane closed before the ball entered it. Pitch control and passing option models do exactly this. They are computationally heavier and they do not fit in a match graphic, so they stay inside clubs.

There is a version of this bias in every sport that logs discrete events, and it always favours the visible intervention over the prevented situation. Deterrence is invisible by definition. If you have a metric of events, and the good outcome is an absence of events, the metric will point the wrong way.

I would like at least one publicly available number that captures denial rather than action. Even a crude one. Passing lanes closed per ninety would do.

Until something like that exists, remember what you are looking at when a defensive dashboard shows a low action count. It might be a passenger. It might equally be somebody who is quick enough that the ball never comes.