Football analytics faces a limitation that no improvement in data collection can remove. The sport simply does not generate many of the events that matter most.
Goals are rare by design
A typical match produces a small handful of goals, and a full season for one team produces a modest total. Every conclusion drawn from goals rests on that small count, and there is no way to observe more of them without playing more football.
Rare events carry large proportional variation, so a team's goal difference across a season contains a substantial component of chance.
This is not a data quality problem and cannot be fixed by measuring more carefully.
Shorter seasons make the problem worse
Football seasons contain far fewer fixtures than several other professional sports, which reduces the evidence available for every question asked of them.
League tables are therefore noisier than they appear, and the gap between adjacent positions frequently falls inside the range of chance.
Conclusions about a manager's impact after a partial season rest on evidence that would be considered inadequate in most settings.
Intermediate metrics were the response
Shot-based measures exist largely because shots occur far more often than goals, which gives a larger sample from which to estimate underlying performance. A season's shots run into the hundreds rather than the dozens.
Moving further down the chain to passes and possessions increases the count again, at the cost of a weaker link to the outcome.
Every step in that direction trades relevance for stability, and choosing where to stop is the central design decision in football metrics.
Player-level questions are hardest
Splitting an already small dataset by individual leaves very little for each one, particularly for players who feature intermittently. Team-level questions are comparatively well served by exactly the same data.
Squad rotation, injuries and substitutions all fragment individual samples further, and a full-back may accumulate only a few hundred defensive actions in a season.
Conclusions about individuals therefore need either several seasons or considerable humility about how much a single one can show.
Regression toward the mean is the working defence
Shrinking extreme observations toward a group average is the standard technique for handling small samples, and it improves prediction reliably.
The practical consequence is that the best available estimate of an outstanding season is usually a somewhat less outstanding one.
Clubs that fail to apply that correction end up paying for performances that were partly chance, which is one of the most consistent and expensive errors in recruitment.