Last week I wrote a big sermon in the wake of the US election results on the future of analytics in football punditry. By analytics I really meant ‘non-intuitive match data’, not necessarily statistical analysis, although that’s obviously a crucial component.

My point was not that statistics will eventually transform football into a mere odds game, but that the time will come when journalistic cliches will have to subject to unassailable empirical data.

What might that mean in practice?

Ideally, it would mean journalists to could point to a number of metrics to make definitive claims about how a team played a football match. A lot of this sort of thing happens already. For example, many bloggers will refer to a pass completion graph to illustrate the effectiveness of a full-back in assisting in attack. This is still pretty crude, but nevertheless is a major step forward from ten years’ ago when the great unwashed had only the word of “experts” to go on.

There is also no guarantee football analytics will be able to do something revolutionary like explain in concrete, tactical terms the cause of the distributive gap between shots, shots on target, and goals. That’s the holy grail for researchers, but in practice, non-intuitive statistical correlations aren’t nearly as important as data that confirms or denies our own subjective impressions.

In an interview from last August, Borussia Dortmund coach Juergen Klopp spoke to his own use of analytics. When told for example the pass accuracy percentage and number of touches of Mats Hummels, Klopp responded:

I know from the trend that was the case and a really terrible pass I would remember anyways. The numbers confirm one’s own impression. That’s all it is. In the long run they are very important because they’re more exact/accurate than one’s own memory.

Klopp should know. He admits to a rote approach that eschews books and UEFA seminars for exhaustive video review, spending 5-6 hours rewinding and fast-forwarding a ninety minute game to diagnose a particular problem or worrisome trend (Klopp reviews a hardly credible 10 games a week). The data simply provides another empirical layer to shore up an impression taken directly from a game.

Klopp was also quizzed on the possibility of future metrics that yield non-intuitive conclusions, like the likelihood goals will most likely be scored from in-swinging corners. He answered:

I could have told you that. These insights don’t tell me much. Because I must devise a plan based on the abilities/capabilities of my players and not on the law of statistics. Perhaps, I come across as open to starting new developments, but in reality I’m still very old school.

This isn’t epistemic closure. We’re no where near the point where managers can reliably tell players to follow through on X action in Y situation and expect a result at a particular point in a single match. Statistical trends are currently at best sign posts that point to general trends against a given set of opponents playing a particular style.

Managers meanwhile have a limited set of players with their own skills and foibles. Coaching is more a matter of plugging holes, correcting recurring errors, and finding a formation and style that best suits his or her players against a particular opponent.

So when it comes to data, what is most useful? That which tells the coach whether a player is broadly playing to expectations, and that which tells the coach if his players are at peak health. Klopp for example when asked which data he relies on, answered “Tempo/speed is important, Stamina-performance abilities and time it takes to regenerate. Moreover, all data concerning health because we don’t want to take any risks.”

As Orwell once aptly wrote, “To see what is in front of one’s nose needs a constant struggle.” As soccer analytics has yet to move beyond its infancy, for now that struggle is enough.