A quick housecleaning note: it’s my intention to get back in the swing of things with a post a here a week. Ideally, I’d like to cover the latest public football analytics work, in addition to other news relevant to better and more efficient front office management. So apologies if this week’s selection is arbitrary, or if I left anything out.
Mostly for work reasons, over the past few months I’ve taken a break from regularly reading public analytics work. There were a few other factors, too. The output is less voluminous, and the work more concentrated in a handful of familiar areas as some have either dropped out of the game, or moved into more proprietary work.
Moreover, since Marek Kwiatkowski’s clarion call for a “new kind of analytics” in August of last year, it felt as if the field would move swiftly into a more technical, more academic or proprietary and perhaps more specialised field, and—despite his claims there is much that can be done with existing event data—one that would rely on expensive and difficult-to-acquire X,Y player tracking information.
As Marek wrote then (and apologies for the lengthy quote):
Almost the entire conceptual arsenal that we use today to describe and study football consists of on-the-ball event types, that is to say it maps directly to raw data. We speak of “tackles” and “aerial duels” and “big chances” without pausing to consider whether they are the appropriate unit of analysis. I believe that they are not. That is not to say that the events are not real; but they are merely side effects of a complex and fluid process that is football, and in isolation carry little information about its true nature. To focus on them then is to watch the train passing by looking at the sparks it sets off on the rails. The only established abstract concept in football analytics currently is expected goals. For good reasons it has become central to the field, a framework in its own right. But because of it focuses on the end result (goal probability), all variation without impact on xG is ignored. This focus on the value of a football action or pattern rather than on its nature seriously undercuts our understanding of the fundamental principles of the game. Just like isolated on-the-ball events, expected goals tell us next to nothing about the dynamic properties of football.
Here we are many months later and it’s still not clear to me, at least as a layperson, how much the community has responded to Marek’s challenge (and it’s possible I’ve missed it among the heap of posts that have been published over the last six months or so).
But I do know that the groundwork for Marek’s kind of approach has long been on the books elsewhere, including tennis in a paper I wrote about after attending the Sloan Sports Analytics Sports Analytics Conference back in 2016, one that aimed to develop “a dictionary of discriminative patterns of player behavior, [so] we can form a representation of a player’s style…” There are still many exciting possibilities for football.
And yes, there is evidence the field as a whole may be slowly moving in this direction. Thom Lawrence and Ted Knutson published their work on xChains back in January, and though it still centres on assists and goal probabilities as their primary measure of value, it at least offers a more holistic look at possession.
Then, this past February, we had Martin Eastwood’s poster presentation at the OptaPro forum in London on the possibility of using Deep Learning to improve our understanding of ideal decision-making in and around the penalty area. And though his methods are above my mental pay grade, this is a fascinating attempt to quantify and present, in real time, the dynamic nature of possession.
One could easily see something like Eastwood’s app prototype used in conjunction with traditional tactical work, even work as dense and complex as Adin Osmanbasic’s analysis of the press and counterpress, published just last week.
Yet there is still a ways to go. Marek’s post was on my mind too when, this past week, I read Harrison Crow’s breakdown of Atlanta United’s 1-6 thrashing of MLS expansion side Minnesota United. Here’s a game in which Atlanta United clearly dominated, but which produced a combined xG tally of 2.04 for Minnesota, and 1.79 for Atlanta.
Crow’s video clips hardly suggest that Atlanta’s well-worked goals were “lucky,” at least in the conventional sense of the word, but I’m not certain the xG model here takes into account opposition positioning (or the weather). In all three cases, Atlanta’s forwards were clearly well ahead of the play. He acknowledges as much:
Sometimes with goal scoring it’s… well, I don’t want to call it dumb luck but, maybe, just circumstances of the moment: weather, keeper positioning, defenders and just the general feeling of that moment. There are tons of variables that we, and others, don’t and in some cases can’t take into account that factor into the equation of scoring a goal.
But at some point, if your aim is to build something beyond what is essentially a long term betting model, you’re going to have to go deeper. Is it really luck or a collection of countless variables that saw Atlanta break so effortlessly? Or was it a combination of the weather and an undisciplined high defensive line? As Crow writes, “Expected goals isn’t sentient and it doesn’t know everything about a game,” but they’re going to need to know a bit more if they can bear the weight of responsibility analysts are intent to give them.
Moving away from analytics and into more ‘common sense’ front office matters, I enjoyed this post by who I presume is Dan Weston, on the weird presumption among former players they have a ‘right’ to fast track through coaching qualifications once their careers are finished.
Let’s think of it like this. Imagine you are the owner of a business and you need to recruit a new CEO. Would you give the job to somebody who has a proven track record of success for a rival company, or perhaps a younger, ambitious director of your company who has demonstrated that he has the potential to step up? Or – would you hire the person who has a great track record on the production line or shop floor?
I think it’s naive to think there are no potential advantages in being a former professional when it comes to management, like a knowledge of football culture, of the rigours of the road, of what’s required to push you forward when you run into fatigue toward the end of a match.
But the analogy I tend to draw when it comes to football managers is orchestral and choral conducting. Most legendary conductors aren’t expected to know how to play every instrument, or how to sing. Some of the good ones will defer to their players on certain matters, like section leaders.
Yet the value in the conductor, or indeed the manager, is that they are a degree removed from the playing; they can see the big picture. As with football managers, I’m agnostic on how much value conductors add to a decent professional orchestra (I’ve met a couple of charlatans in my day), but it is absurd to think conductors must have experience playing the instruments to be effective. They just need to know everything about music. One could easily say the same for football managers.