Just run some more

Fashions come and go and, inexplicably, 90s aesthetics are back and running stats are a hot topic.

Well, maybe it is explicable.

Football has had regular running data since the 90s, but a few things might’ve held back its usage and usefulness. One of them is ball tracking - harder to do than player tracking but really helps to know what’s going on on the pitch. Another is arguably imagination (see ‘Are Current Physical Match Performance Metrics in Elite Soccer Fit for Purpose or Is the Adoption of an Integrated Approach Needed?’ (2018, Bradley and Ade))

Once you have that, it’s easier to chop the numbers up by phase, and when you can chop the numbers up, they get more interesting and useful. Split them by position and tactical phase, and you get a very quick, very nuanced picture of a player’s physical demands on matchday. As I often do, I’ll point to the FIFA Training Centre as an example of this kind of work.

That detailed matchday information also lets you work backwards. Backwards from matchday: to training, to recovery from injury. MLS team Chicago Fire are using French company Footovision to help with this. Other systems will, one presumes, be available. (Perhaps like fellow French-born company, Skillcorner).

Platforms have power. It’s not just Parisian-based companies springing up over the last few years: over the last couple I’ve noticed a few research papers using data from the Kitman Labs platform (like this one on hamstring injuries). My venturing into the sports science realm doesn’t even reach ‘amateur’ status, but I’m familiar with the ‘n=8’ sample size problem. This paper, though, is able to call on data from 36 team-seasons in elite men’s football.

Software (usually) means scale. Loop back to the FIFA article I shared earlier: that was about women’s football, data from the 2023 World Cup. As the sport tries to catch-up on its historic lack of support and active hindering of women’s football, some things scale better than others. Training infrastructure and coaches can only be shared so much. Computer vision data collection scales better than event data collection. The insight into how to split tracking data into tactical phases, to benchmark players against their peers, barely needs to scale at all. Once it’s there, it’s there.

And the global game being the global game, there are geographic implications. Check out the countries that viewers were tuning into this computer vision for tracking data Q&A stream from - there’s Turkey, Uzbekistan, India, Congo, Uruguay. I won’t repeat my previous line about ‘why not make your own tracking data instead of buying it’ (also from that September ‘22 post above), but a rudimentary system may still have value. Even if you work with large error bars, maybe you could take inspiration from the literature, gather non-quality assured data for the Uzbek second tier, and be able to benchmark attacking full-back requirements against their peers.

So, yeah, running data is cool again. Ditch the 2010s slimfit jeans, dig out the colourful maximalism.

There’s another advantage that running data has. No-one has to argue about what a ‘duel’ is.