A post-conference post

Hard work is a necessary evil, until it's not. That is, I think, my pithiest thought coming away from the Hudl Performance Insights conference this Friday.

Coaches used to spend hours creating analysis clips, before services became available to do that for them. Years later, data scientists (in the broadest sense of the word) would need to put messy tracking data in order, warehousing it and smoothing it and then processing it into manageable chunks. To a growing extent, this is now being done for them by data providers.

Of course, a lot of these time-savings are not eliminating hard work, merely displacing it somewhere else. In the early days of Prozone (which had a nice panel at HPI), it sounds like an incredible amount of work was done to get the early tracking data processed. Manual event data collection is still obviously quite labour-intensive, and making raw 'sensor' data suitable for use is not any easier just because the data company is doing it instead of someone at a club.

However, by centralising the work, accumulates a lot of labour displacement across each of the customers (and technology has made some of these tasks easier in and of themselves).

I was struck, though, by anecdotes in a panel about Prozone's early days, in the early 2000s, that sounded very like early 'analytics' history in the early-to-mid-2010s. Things like the need to get 'advocates' on board in clubs, often younger assistant coaches or analysts; the effect of big-name head coaches as users, who their peers wanted to emulate; the potential for the board to use the data to have their own benchmark of the manager's performance; the resistance of some within the industry who were either sceptical or just cagey about the change.

And there's an echo of this with generative AI-powered tools (in and outside football) a decade on. It's quite common, at present, to hear people say things like 'there's value in the friction' of doing work yourself instead of offloading it to AI, and I'm sure that if that phrase had been coined in the 2000s it would have been used as an argument for watching matches in full, live or on tape, instead of looking at data or clips from Prozone.

It's certainly true that there are tangential benefits to watching a full match rather than the clipped or data-ed version of it. Just like the happy accidents of creativity, an important detail could be found while watching the game for something different. But it's also simply true that, when wanting to specifically know how X does Y, it's more efficient to just look at instances of X doing Y.

It can be difficult to put your finger on exactly why you might prefer to do something yourself, however 'inefficiently' that may be, rather than offloading that work. The happy accidents might be part of it, learning is also certainly part of it, but I wonder whether the current psychological/emotional responses to AI might also be able to be applied back to those coaches encountering those technologies for the first time. A sense of a loss of control; or a loss of identity, if 'the grind' was something they particularly prided themselves on.

There's a curiosity amongst all this though.

The closing session of the day was a conversation with Gaizka Mendieta. He talked -- amongst other things -- about how he did extra training sessions to work on his technical ability, eventually changing positions from a holding midfielder/right-back to an attacking midfielder. He also alluded to the fact that players tended not to like watching the video that managers gave him, in a tone that sounded a lot like a stereotypical child describing stereotypical homework.

But Mendieta, and presumably all of his peers, did not shy away from work and learning; a stat-sceptic coach in the '00s or 2010s who watched tape all hours of the day wouldn't have feared studying. What, specifically, was it about players watching video or coaches learning to use technology that jarred with them?

For a long time, I think high-level professionals in football were justified, to an extent, in being wary of data. As Mendieta (and many, many others) have said, stats on running and passing need a lot of context and could easily be mis-used. They didn't describe the game in all of the sport's sophistication.

Nowadays, though, data sources are getting very close to doing just that. One of the points in my talk on the HPI Research Stage was that, over the past few years, there's been a growth/improvement in the ways that tactical and physical load concepts have been represented in data. Not only are the metrics catching up with the ways that coaches speak football, they're arguably verbalising it better. The language is getting more precise. The vocabulary is getting broader, often more specified, and -- because the work is being centralised within data providers -- more unified.

There are a bunch of different parts of a non-playing football job where there's a lot of hard work. A bunch of things will still be done relatively manually, and there are a lot of decisions to make in uncertain environments. But is there any of that hard work that is not, in fact, a necessary evil? Because sometimes things can just come along to lighten the load (usually to be filled by a different type of, hopefully valuable, work).