I’m often told that once you get into gambling, there is no such thing as a boring football match. Though I am too scatterbrained and busy to properly make bets on games myself (it is also a journalistic hazard), I can understand the idea.
What to the neutral might be a dismal, grinding, boring mess of a match will to a gambler be a series of discrete events that, in aggregate, will lead to an statistical likelihood, one that, should it ultimately come to pass in the way they predict, will earn them a small profit. The bettors enjoy sports, of course, and they may even appreciate its batshit insanity from time to time, but for them the thrill of the game is not from its intrinsic beauty or captivating story-lines but from predicting its final destination.
This past Saturday saw a rare awful Champions League final, a final worthy of a World Cup, between Liverpool and Tottenham. Jokes about all-English match-ups aside, it didn’t get the neutrals’ hearts racing. I was on an aircraft with expensive wifi that didn’t support video streaming, so I missed almost the entire game. From what I gather, I was spared.
Shortly after, Jonathan Liew wrote about the boring final as a reminder that despite the need of the sports many promoters for the sport to be entertaining above all else, football—despite its rising star-power and rising goal per game stats—it is still a a sport unwilling to perform on command. Teams set out to win, not showcase the beautiful game. Weather doesn’t always cooperate. Players get tired. Boring matches are a part of the game, despite our belief we can legislate them away (say by moving back to a 16-team World Cup or getting rid of the group stages). Liew writes:
Perhaps that, despite all the attempts to harness it, football at its core remains defiantly resistant to control. That despite every imperative to flatten out the peaks and valleys of sporting chance in favour of smooth business certainty, once the whistle blows you can’t guarantee a thing. That the biggest and most lucrative sporting fixture of the year, the pinnacle of the men’s domestic game, was a total washout. I don’t know about you, but I find that quite cheering.
For the professional better, however, ‘flattening out peaks and valleys of sporting chance’ is required for their livelihood.
Sports bettors are a critical constituency in sports analytics; in their drive to beat the betting lines, they developed innovative predictive tools using sometimes the most basic of event data, many of which are now in use by club analysts wanting to better separate random variation from underlying performance.
Of course, the bettors’ wares also found a mainstream audience on self-consciously data-driven sites like Nate Silver’s FiveThirtyEight. This includes the win probability metric, the subject of a fantastic piece on The Outline by Jeremy Gordon this past week.
Like Liew, Gordon believes that win probability cannot capture the ineffable, linear dynamic of a lived out sports contest in real time:
This ongoing emphasis on the statistical likelihood of a given event taking place is just sort of useless, in the final run. It turns sports — a great form of entertainment predicated on watching the most talented athletes alive do incredible things with their bodies, like this crazy shit — into a number-crunching exercise, in which the competition is formally understood as a series of distinct and modelable events culminating in the final score.
The broader point here to me is less about win predictability being a boring and overly reductionist way of understanding the maddening ebb and flow of individual games (it is), but about the problem of using things that were primarily derived from bettors tools for ‘entertainment purposes.’
For example, some sports media companies that think enhancing sports journalism with analytics means adding numbers in your match report for ‘context.’ In practice, that might mean something like saying the win probability percentage of a game was this, then this thing happened, and after it was this.
This information might be worthwhile to the tiny sliver of fans who make in-game bets, but for a broader audience, they can feel either soulless or, for those for whom there is not a sports-related number that isn’t accurate, be a blunt tool to make bad arguments.
To me, it’s also bad analytics. For one, it feeds into the persistent myth of the single match outcome as predictive and deterministic. Whether we’re talking about win probabilities or xG totals, the result is the same. By totaling these, we are tempted to make lazy assumptions about what should have happened in a game driven in large part by random variation.
For another, it’s a dangerous crossing of the streams. It’s one thing to adapt sports betting models to help provide an accurate picture of whether a team is suffering from poor luck or is overperforming etc., and it’s another to take a metric designed to help improve the likelihood of getting a return on a bet and using to make qualitative statements about a team.
I’m just asking that we not be lazy in how we use numbers in sports journalism, and my issue fundamentally with things like win percentage is that they are lazy. In the same way we cannot force football to be entertaining, we cannot force it into our narratives, nor can we fence it in with inscrutable numbers based on discrete events.
I’ve said this before, but I think the key is to just tell the story of the game as best you can, try to analyze tendencies as best we can, and do so in the wider context of how a team has generally done over a longer period of time.
But just like the universe becomes increasingly unpredictable, weird and inscrutable the smaller the scale you view it through, so too does football. Ticking probability lines don’t help. Neither does getting angry that the game isn’t living up to the media-generated hype.