What, ultimately, drives an amateur stats analyst to first try their hand at football analytics?
Is it the idea of building a career, the prospect of becoming football’s Paul DePodesta, revolutionizing the way the game is played?
No. Ultimately, I think, that comes later when the analyst realises their ideas are gaining some traction.
Is it to make money betting money on football?
You will find that there is a silent network of individual betting modelers who don’t talk about what they do, because there is no need…in fact, it’s counterproductive. The desire to publish work comes from a different place.
I think the core motivation for any analyst is the belief that they can help football clubs do better, whether that means not sacking the manager out of panic, not wasting bags of cash on overhyped, aging stars, not crossing the ball 60 times in a match, whatever. For better or for worse, most analysts start publishing because they think they have a unique set of skills that can help teams improve on what they do.
If that’s true, then analysts need a model for the best way they can provide that help, beyond just pressing publish.
So, an analogy. Imagine for a moment a friend of yours is struggling with something at work, in an area that you strongly believe you can offer meaningful help. The only problem is that, while you’ve thought a lot about their specific set of problems, you don’t have much personal experience helping others in a similar situation, and only a few examples of where similar assistance has worked with other people facing the same issues.
What do you do?
Perhaps you decide to tell your friend firmly, “You’re doing this wrong. Let me show you why you’re doing this wrong, and how trying this might help.” So you sort of step in take over, detail convincing theories on how everything will work out in the long run if your friend just trusts the process. When they ask you if this has ever worked before, you reply, “You’re going to be a pioneer.”
Things may go very well, but without a feeling of partnership, your friend may be inclined to blame you and your methods at the first sign of trouble.
Maybe your friend asks you to coffee so you can help them out, but when you show up, there are three or four other people there, each with different outlooks and perspectives on the same problem. Though you’re a little miffed you’re only one voice of many—particularly as you feel the other friends are offering poor advice—you nevertheless listen patiently and wait your turn, raise your objections, and detail them clearly. Some are acknowledged, others discarded, and your friend decides to make a plan of action that incorporates a little bit of all the advice they’ve received. Afterward, you tell yourself you did your job and you should be proud, even though you’re skeptical out the final outcome.
Or, maybe you and your friend sit down one-on-one. You start by saying, “Describe in detail what it is you’re struggling with.” You listen for a long time, and when you’re done you might ask, “What is it that you feel you need help with?” And you do some more attentive listening. After, you resist the temptation to start rattling off solutions but instead go away, take some time to think, and come back with a few ideas you think might help based on what you’ve heard. Maybe your friend likes some solutions but thinks others are impractical; you politely ask why, and then reach a compromise. Then you ask your friend to try the ideas out, and track what works and what doesn’t.
The Importance of Story
Before I come back to this analogy, let’s look at a few thought-provoking articles on sports analytics this week.
The first, from 538, talked about broad growth of the analytics field in Major League Baseball by Ben Lindbergh and Rob Arthur. There were a couple of interesting highlights from this. First, that the data suggests it paid for clubs to get in on the “first mover” advantage in baseball analytics; even a five person analytics team in the early going could add a couple of added wins a year, worth a substantial sum when compared to players that offer the same advantage. The ROI on analytics is substantial.
The other is that, counter to what many initially predicted would happen a decade ago, traditional scouting has thrived in the analytics era. Part of this is because teams have worked to increase their international prospect pool, which requires putting asses in the seats in increasingly far flung locales. But the other reason is that scouting perspectives offer an added scope to raw data:
If anything, smart teams have learned to treat scouting grades as statistical data that can improve upon purely numbers-based evaluations, making the two perspectives even more tightly intertwined.
Yet there is another development captured in the data which the authors glossed over a bit.: clearly, some Major League Baseball teams believe more is better when it comes to their analytics departments. The authors take it as a given that “bigger R&D” departments mean clubs are “more committed” to analytics, which is probably true. But this raises an interesting question: is bigger necessarily better? Do more analysts produce better results?
At the Sloan MIT Sports Analytics Conference this past March, I spoke to someone from a leading data firm who wondered if bigger analyst teams on football clubs were a good thing. After all, a bigger analyst team might mean its individual members have less ownership over their work, less of a sense of mission, less of a chance at integration with the coaching staff, particularly as the team grows into a separate “department.” Perhaps Leicester City’s comparatively “lean” approach to analytics might pay greater dividends in the long run, where analysts aren’t add ons, but vital, integrated parts of the team, part of the winning story.
This is an interesting but untested theory, and it leads nicely into the next article by Sam Miller, on an analytics experiment in minor league baseball in the New York Times. Interestingly, one of its subjects is the co-author of that 538 piece, Ben Lindbergh:
IN 2015, the Sonoma Stompers, the team with one of the lowest payrolls in the Pacific Association, a professional baseball league near San Francisco, did something desperate: It handed its baseball-operations department to a couple of stat-savvy writers with no baseball-management experience, Ben Lindbergh and me.
Though things went swimmingly in the first half of the season, the Stompers went downhill in the second. The reason, Miller theorizes, is the Stompers didn’t have a compelling story:
We sold our story as something imposing — “data analytics” — and we made it about us. We should have sold it as providing them information, and made it about the team. That would have fit into their view of the sport — that we were trying to give them the same resources major-league players like Miguel Cabrera and Clayton Kershaw get. With other sabermetricians, more data wins arguments. In the dugout, a good story does.
You might think this is newspaper narrative bullshit, but Miller has a point that I think is vital to understanding how analysts might view their role in the game.
In his recent book Better, Smarter, Faster, Charles Duhigg talks about the psychological importance of the feeling of being in control when it comes to motivation, obviously a crucial part of what it means to be a winning team. Yet even more important than this feeling of control, writes Duhigg, is the idea of meaning. For players to want to win, to do everything they need to do to maintain focus on the end result, they need to feel they’re all helping to write a compelling story.
So if, for example, as a player you feel like a pawn in some computer nerd’s experiment, you may not feel as personally compelled to compete with the same fervour. However, if you’re an underdog striving to help out-think your rivals as part of a risky, unconventional approach, you might be far more committed to the end result. And that doesn’t go just for players but analysts, coaches, executives, directors…the entire organisation.
But I also think it’s hard to maintain that sense of being part of a story when you feel like you’re one, tiny cog in a huge organization, whether on the field or in the front office.
Which brings me to Ted Knutson’s great article on the inevitable stats revolution in football over at StatsBomb. Knutson too wants analysts to start thinking like helpers, not just educators. He writes that it’s not enough for analysts to apply statistical analysis to football data to “learn new things’; rather, they must work harder to demonstrate how their methods will give clubs a greater competitive edge. He gives some examples:
How can teams compete with the traditional giants beyond just spending more money?
Apply the marginal gains.
Make consistently better decisions than other teams.
Play more efficient football.
Recruit better coaches.
Recruit better players.
Make fewer mistakes in the transfer market.
Find. The. Edges!
Knutson explicitly mentions the Premier League here, and I think football offers another, potentially unique opportunity for smart analysts to get involved.
Analytics and Supporters Trusts
Sometimes I imagine what I would do if I were a proper football analyst and some poseur Canadian hack.
One thing I might do would be to take some time to visit a number League 2 and Conference clubs across England, ask to sit down with people in the front office, whether managers, board members—who was willing to talk to me, basically—and ask: “What are your biggest challenges right now? What do you feel you need help with to improve, beyond just a giant bag of money? Where do you want the club to be in five years time, both competitively and financially? What do you think it would take to get you there?”
I would take notes, find similarities, differences, and then take some time to think to myself: what, for these clubs, is the low hanging fruit that an analytical approach might help with? What would be the challenge to me as analyst working to help these clubs (lack of data would be high up there)? What innovative solutions could I come up with to help address their unique problems?
Then I would return to these clubs and talk about some of these ideas, and I would listen to their responses. Which of these ideas is practical? Which do you foresee having trouble implementing? Which ones are you most skeptical about? Why? What would it take for you to be willing to experiment with a different approach?
Returning to the friend analogy I opened with, the point is not to come in and say, “I know better about this than you.” Rather, it’s to listen, to offer help, to be a partner, not an expert. And I think that’s something that’s potentially far easier to do (though less lucrative) at a smaller team than at a bloated Premier League club.
Of course it’s possible that football’s closed, conservative culture is far more entrenched further down the pyramid, and making inroads here would be just as difficult. But that brings me to the other part of this fantasy—compiling this information into a document to share/co-author with Supporters Trusts.
As more and more fan groups vie for club ownership, STs face a challenge—how to bridge the gap from supporter to owners? I passionately believe analytics should be at the heart of this conversation.
For one, a partnership with supporters trusts might help dispel the myth of the analyst as “bean counter,” working on behalf of stingy first division owners looking to win on the cheap. And by aligning themselves with football’s grassroots culture, analytics can make credible in-roads in the game, not as an “army of geeks” but normal football supporters with a set of specialised skills that could be of help to smaller clubs.
I also think that analytics can be more effective in smaller organizations where the margins of error aren’t always as catastrophic, but where finance is an ever-present concern. It also poses a unique challenge for analysts—what novel ways can you overcome significant gaps in event data? And how can you push for better data at this level of the sport?
Which brings me to the question of money…could the analyst make a living from this kind of work? I don’t know…probably not in the short term. They could make a name for themselves, however. But I think it’s still worth pursuing, in the same way I think that the core motivation is the same for the analyst who first decides to hit publish: they want to help.
Don’t want to take too much time, but there are two articles that partner well with what I wrote last week on Expected Goals: Martin Eastwood’s take on the importance of uncertainty, and Garry Gelade’s look at how passing before shooting can help raise shot probability. Both fantastic pieces of work.