Soccer analytics in the 2020s
What's it gonna look like? What's going to change? What should change?
How far we’ve come in just a decade
Before I get deeper into this, I have to confess that my involvement in the soccer analytics scene dates back only to 2011-12, give or take.
That’s when I started as a feature writer at a media company that would eventually morph into theScore.com. My job when I arrived was, like a lot of other internet writing jobs earlier this decade, about pumping as much content online as possible. So something like 6-8 300-500 word blogs a day.
Football is, thankfully, a sport that tends to lend itself to that kind of insane production line. But my problem is that I couldn’t sit there and rehash leading news stories of the day (though there was a lot of that). I needed the stuff to be somewhat novel, beyond my poor attempts at trying Deadspin-esque humour (this is where I’m slightly thankful that theScore nuked all my pre-2014 content).
The inspiration to start poking around the soccer analytics scene came from my co-workers, guys who were/are fantastic sportswriters in their own right: Scott Lewis, Drew Fairservice, Dustin Parkes and Andrew Stoeten.
This was a crew who was aware of and receptive to what we mostly called ‘advanced’ or ‘fancy’ stats, but also skeptical, and not in an old school newspaper hack kind of way. That alone to me was unique—to be savvy enough not to dismiss analytics at first glance but without the need to dedicate their waking lives to caring about things like Corsi and PDO figures.
Even more interesting to me was that, while there wasn’t a ton of public soccer-focused analytics work that I could find, the intriguing independent stuff owed a lot to ice hockey. One of the first guys I stumbled on was James Grayson, who was preaching a shot-based stat, now out-of-vogue, called Total Shots Ratio, as a crude way to predict future performance.
Then over time, I discovered Ted Knutson’s old blog, Mixed Knuts, and people like Sarah Rudd at StatDNA and Ravi Rameneni. I went to my first Sloan Sports Analytics Conference in 2013. I got in touch with some very nice and smart people in Canada, the U.S. and overseas. It’s been a fun eight years!
Things have obviously changed quite a bit since I started learning about this stuff. Ted is now the head of a fledgling analytics company. A lot of the dudes I used to Tweet at for feedback on blog posts now work for major European teams. I work in an office job unrelated to sports *cough*.
More importantly, there is strong indirect evidence that clubs are living the Gospel of Expected Goals, taking higher value shots from more dangerous positions rather than cough up possession in the class hit ‘n hope—a remarkable development driven in part by data analysts:
But despite these advances, part of me feels things on the metric/model development side have been more or less static/stable since 2016-17. Though it’s an engine for a lot of interesting models, xG, or at least the overall concept that drives it and its myriad offspring—averaging the effectiveness of certain on-field actions to judge its value on the pitch—is still in the driver’s seat.
That’s not a bad thing. But it’s also not the Promised Land those in the analytics universe have been discussing practically since 2013, pushed by geniuses like Luke Bornn: tracking data.
Will the 2020s finally see tracking data drive accessible and actionable predictive models and metrics?
Interestingly, a recent, widely shared soccer analytics blog post comparing two of xGs conceptual offspring illustrates the current paradigm. In both cases, we see an attempt to ‘workaround’ the lack of positional tracking data:
The contributions that we can measure are inherently limited by the data we have. This post focuses on approaches for event stream or play-by-play data. That implies that the data only contains the location of the player possessing the play. The locations of all other players is unknown. […] While obviously limited, this data is much more widely available than optical-tracking data.
I know some analysts who have worked diligently (and patiently) with both companies and their clubs to begin to compile tracking data but getting there is no mean feat. Paying for cameras that track this stuff involves a lot of trust from the league and club, and the big question—one that I have yet to see a satisfactory answer for—is how timely and actionable will this data be?
That’s probably a question someone like Bornn could answer. And, of course, it’s more than possible clubs have already gone far down this rabbit hole, and the public hasn’t heard anything about it yet. But as of 2019, the promise of tracking data to produce exciting new metrics and models that will forever trump the old counting stat-based stuff has yet to materialize.
Analytics alone isn’t enough
Though I’m completely and utterly biased, one of my favourite ‘analytics’-based companies has always been 21st Club (I used to write for them).
The London-based company headed by former Opta guy Blake Wooster and staffed by a group of brilliant people go beyond merely providing and developing metrics or software to clubs, but encouraging, where necessary, a wholesale culture shift. They aim to move clubs away from thinking about short-term, reactionary decisions toward planning a coherent, top-down multi-year strategy that takes into consideration the possibility of failure.
Though rarely talked about, football culture is permeated by fear at all levels. Fear of relegation. Fear of administration. Fear of the TV bubble bursting. Fear of fans. Fear of not making the first team. Fear that expensive transfers will be duds. Fear of getting sacked as manager. Fear, fear, fear. And as the revenues in football have grown, so too has the pressure to succeed, and the fear of failure along with it.
Though analytics is often framed as a means to gain a competitive edge in sports, it’s also a tool to combat fear, which often comes from uncertainty. Though not perfect, analytics can help give you greater confidence that your transfers won’t flop (though the public record is mixed here) or that your losing streak is down to bad luck and you don’t need to sack the manager you just the year before.
But analytics alone isn’t enough. Clubs need coherence. They need to agree on an overall approach, they need to get the right hierarchy in place, they need to ensure their player academy shares the same values as first-team coach, the manager and the director of football. They need to hire managers to fit their overall club philosophy, rather than write it every two years.
Even the most analytics-savvy teams struggle with these issues. But this year especially, I think we’re finally getting a sense of what a coherent club might look like in Liverpool, a team that, despite being favourites for this year’s PL title, is already thinking about the next five years and beyond. Let’s pray they’re not the only ones as we head into the twenties.
Wither fun, free analytics work?
I’ve been on the record in the past lamenting the lack of fun, new analytics work from chatty, argumentative people who don’t care about things like intellectual property and privacy and black boxes and all that shit people talked about back in 2015 or whatever. But that work still exists, except it’s primarily been absorbed into the analytics conference circuit. I.e., there are still a ton of exciting takes and papers flying around, they’re all just connected to things like the OptaPro conference or Sloan or the myriad others we see every few months or so.
Yes, it’s less spontaneous than when people would post some interesting shit on SB Nation team blogs, and then others would respond vehemently on blogspot.com blogs. One thing I do hope to see in the twenties, however, is a move away from the traditional academic presentation style and into applicability. Or, at the very least, maintaining a dual approach, where you present your case for a particular model or metric, and then in a follow-up post, try to make clear in detail how a club might theoretically and realistically apply your metric or model to improve the side.
We still lack this element, and it’s crucial, absolutely crucial, to get more teams on board with the potential of these metrics to give clubs a competitive edge.
Happy holidays and many thanks to you all
Just a note to say that I may be posting a little less toward the end of December as I deal with holiday shit with the fam. I also wanted to thank all of you for supporting this newsletter, my last tether to the still fascinating world of football analytics.
Finally, I wanted to use this end-of-decade to say thank you to the countless people who I’ve been able to talk with, work with, laugh with and agonize with when it comes to soccer analytics. I’m incredibly privileged to have been able to meet you, even when you’ve said nasty things about stuff I’ve written.
All the best to you!