One of my happiest professional memories comes from my final month working at theScore in July 2014, where I was a “features writer” for three years.
I had spent most of the 12 months before writing about the burgeoning football analytics’ field, and the implications for how we understand the game. This meant trying to think clearly about games, title races and how teams played without peddling the same cliches you find in sports writing just about anywhere.
My final assignment however was to help cover the 2014 World Cup in Brazil. Here was a tournament that was largely inscrutable from a statistical point of view; international sides only played a few games a year, and were now going to engage in a month-long knockout tournament. As I wrote for Foreign Policy at the time:
Half of the 32 teams will play a maximum three matches, and only four teams will play the maximum of seven. Their opponents depend on the initial draw and the final places in the group stage. One team could face France and Germany on its way to the final, while another might play Mexico and Greece. With such different opponents, comparing the performance of teams or players would be all but hopeless. And match data from this tournament, no matter how interesting, will almost certainly be skewed by unlikely events — the kind that can turn a single game but might never repeat themselves in the course of a season. If none of the data is repeatable, it’s hard to see how it’s descriptive.
This may sound dismissive, but it actually gave me license to return to the the kind of football writing that first made me, well, want to write about football—I’m thinking here of the Two Brians, Glanville and Phillips. Writing that, I hoped at least, would eschew explanation in favour of celebration. And so for theScore I wrote free wheeling stuff like this:
The semis, meanwhile, provided two contrasts in extremes. One involved Germany’s mindbending 1-7 destruction of Brazil, a result that is still too big for any of us to quite grasp yet, but may have already permanently wrecked the Seleçao’s Nike-approved status as a world footballing power still capable of Garrincha-style sexiness.
The other match was the nadir of the last 28 years of big tournament futility—two sides not willing to engage out of fear of failure, fear of conceding, fear of losing, rather than love of winning. One wishes Argentina or the Netherlands—two nations that produced Maradona and Cruyff, and now Messi and Robben—had taken the advice of 1986 World Cup winner Jorge Valdano:
People often say that results are paramount, that, ten years down the line, the only thing which will be remembered is the score, but that’s not true. What remains in people’s memories is the search for greatness and the feelings that engenders.
This wasn’t calculated or cynical; I meant, and still mean, every word of it. It was, to this day, one of the more fun writing gigs I ever had, and it ended up being my last for theScore; a month later I was laid off along with the other, more talented full time staff writers.
However, to write that way I had to suspend the knowledge that I was telling tales, shaping one off matches between often even balanced teams that involved complex, non-repeatable events into a sweeping story.
This didn’t take much effort. It’s what most of us do when we watch sports—we give into capital ‘m’ Meaning, even though that meaning may be subjective, sui generis, illusory.
Football is Objectively Ridiculous
Nor is this a bad thing; there’s something deeply irrational about football, or any sport for that matter, and that’s part of why we love it so much. I mean, the very idea of soccer is, from the outside, kind of ridiculous. Millions of people form exclusive tribes based on the most arbitrary connections to clubs that pay people a lot of money to work together to kick a ball into a net on grass fields around the country. That any of us would take it seriously is, from an outsider’s perspective, pretty weird.
Statistical science however can’t operate irrationally. Football analytics uses statistical science to ask a very simple question: how do teams win? To work, it can’t have any presuppositions about the answer, like “It’s about X,Y,Z tactics” or “It’s about grit, heart, fight etc.” Note, it can’t also dismiss those explanations out of hand, either.
But if our interest in football is irrational—based entirely on the in-group bias—while statistical analysis is rational—ruthlessly grounded in the scientific method—sports and analytics may be fundamentally irreconcilable. And, you know, that’s okay. After all, what fun is a World Cup when we choose see it only through this kind of lens:
This may come as a surprise that I’m writing this considering last week I argued that, in practice, team analytics should be framed as part of a story. But there’s a difference; the use of analytics, not analytics itself, can be a motivating narrative for players, coaches, managers who need those kinds of stories to do their jobs.
Again, coming back to what I’ve written in the past, I still think the use of analytics is fundamentally romantic. Think of the appeal of stories like Moneyball or The Extra 2%, books on how analysts communicate with managers, books about arguments and in-fighting, about major insights and progress and terrible mistakes, books about how teams manage to beat wealthier opponents using only their wits. Great!
Yet this should ideally remain a one-way street; analysts need to be careful to ensure their work doesn’t succumb to the temptation to fudge, to win converts by uncritically accepting sporting cliches that may or may be accurate.
For their work to be effective, analysts must be as honest as possible, first and foremost to themselves. While they should be very open-minded—there is nothing more insufferable than the sports analyst who lacks an imagination, and I’m looking at you Hockey Twitter—they must also be careful to ensure that open-mindedness doesn’t not devolve into polite deference for popular, strongly-held cliches that are either unexplored or flat out untrue.
Constant immersion in newspaper columns and post-match TV panel discussions can make this difficult. Which brings me to Wednesday’s title bout between Craig Burley and Gab Marcotti on the subject of Bayern Munich’s superior Expected Goals tally vs Atletico Madrid.
This was touted by some as a “debate” on Expected Goals. In reality, it was an ambush, and not simply because Burley jumped down Marcotti’s throat for even mentioning the metric.
The truth is Gab was doomed long before Burley opened his mouth. The segment was four minutes long. The very topic of debate itself was absurd: “Was Guardiola a failure at Bayern?” A failure in what sense? What were the expectations, who sets the criteria? Is it not winning enough trophies? Had Pep made the final but lost on penalty kicks, would he have been a failure then? It’s a hopelessly unclear topic.
Worse still, the debate was tainted by the halo effect. Did Pep Guardiola and not Bayern Munich lose in the Champions League semifinal? One must consider the possibility that Warren Buffett’s wise words on CEO’s may apply to football clubs as well: “a good managerial record…is far more a function of what business boat you get into than it is of how effectively you row.”
Great teams attract great managers, and discerning where team talent ends and managerial influence begins can be difficult. Was it Pep Guardiola’s fault that Thomas Mueller missed his penalty? Do any of us have a strong sense of where the fault lines lay between Guardiola’s tactical preparation, the intentions of the individual players, and random variation? If you do, I strongly suggest you give up your career and either become a full time professional bettor, or get your UEFA B license ASAP.
And were Expected Goals even particularly germane to Gab’s point? It was fairly obvious that Bayern had the better opportunities and looked more dangerous on the night—they won the game, in fact! At the very least, it’s safe to say that Atletico were somewhat lucky to not have conceded a third.
But the tie was won in the first leg, where Atletico—already one of the best defensive sides in Europe—looked far more assured at the back. But even then, how much did home field advantage play a role? And the away goals rule, for that matter?
Ambiguity, probability and uncertainty don’t make for sexy TV of course, which is why Burley’s glass-cutting insistence that “results are all that matter” will have likely rung true for much of the viewing audience. The medium is geared toward “This Means That;” in fact, sports culture itself is driven by the idea that results are all that matter.
Results, not vague probabilities, are what wins trophies, after all. Leicester City won’t have to concede their Premier League title because they had an unnaturally low shot conversion rate against, nor should Claudio Ranieri be forced to apologize for the media awarding him far more responsibility for their unlikely win than he objectively deserves.
Moreover, Craig Burley’s perspective, while obtuse, is a potentially useful mental model, in the same way xGs are part of a useful mental model (pardon the Buffettisms today). It may be better for club managers to speak like Burley and think like Marcotti, lest players don’t strive as hard, comforted by the knowledge that superior performance metrics will let them off the hook for an L.
This is why I’m wary of media introducing metrics into match reports or post-match debate shows, or mainstream sports media in general for that matter. We already see this sort of thing in a lot of North American sports. If you half-ass it—if you take an approach that broadly repeats the sporting cliches which are part of why we love football but with metrics “added in”, not only do you risk seeing meaning in stats where it may not be there, you can also leave the reasonable impression among viewers, who likely don’t know how the metrics are used in practice, that the stats are redundant. “Oh, one team took better chances than the other? Thanks!”
But that’s not what these metrics were built for. They’re for making predictive models, more sharply defining probabilities, making sure teams make educated bets. These models are fundamentally agnostic; they don’t care one way or another if managers matter or if there are truly streaky players. They either work or they don’t.
I may be wrong on this, but I’ve come to believe that, as domains, what makes sports broadly appealing and the process of analytics itself are irreconcilable. If you go all the way down the analytics/cognitive psych wormhole—which you arguably have to do if you want your analysis to be effective—you blow up most of what makes sports awesome to a lot of people. You enter a world where there are no “world shattering” World Cup semifinals, no all-powerful genius managers who oversee every last match detail perfectly, few if any underdogs who wins a title on grit and determination alone.
I’m not arguing this is specialized knowledge either. Deep down, we all know this about sports, in the same that knowing Game of Thrones is fictional doesn’t stop us from discussing the motivations of its characters. It’s not the games themselves that matter, but what we project onto them. Professional bettors earn higher returns when they can see through this added element, when their models suggest the lines are driven by our false projections. But that “added element” is also, I believe, often essential to what makes sports fun.
You can’t have your cake and eat it too. So just eat it.