In a recent extensive piece on the use of data science in analyzing player performance in football (or soccer, for our US readers), New York Times author Rory Smith argues that while the use of data to support team management decisions like transfers, lineups and on-pitch tactics is now widely accepted, acquiring the right data science talent to drive these efforts remains a challenging task for teams.
The article describes the story of German Bundesliga team Eintracht Frankfurt and how their new sporting director hired his old team’s head of data analytics as one of the first steps after his appointment. Although largely unnoticed by media and fans, the move exemplified how rare good and proven talent in the football data analytics field currently is and which interesting challenges the market for such skills faces.
A lot of us remember reading the book or (like in my case as a die-hard Brad Pitt fan) watching the movie Moneyball, where the seemingly disadvantages baseball team Oakland Athletics uses a data-driven approach to scouting and game tactics to go on a record winning streak and disrupt the world of sports by proving that data can more than supplement manager experience. In football, a significantly more dynamic and interactive game, setting up hypotheses, deriving inferences or even predicting future outcomes is as challenging as it gets. And just like in the case of baseball, football coaches and scouts are usually strong characters that consider themselves more than capable of making the right decision based on their experience and game philosophy.
The article points out that even though most teams have embraced the usefulness of player data analysis and even data on coach behavior, the struggle lies in attributing performance to the benefits of player analytics while controlling for other variables. There are no treatment and control groups in football and running experiments by simulating real matches would still introduce noise, not to mention the costs and additional strain on player fitness.
An additional barrier in truly analyzing the impact of player analytics in football is the secrecy of teams on their methods and results. Thus, while a hired data scientists may be the best in the field and create incredible results (think Jonah Hill from Moneyball), the reality is that the public or even other team’s management will likely not find out, or as Smith puts it:“Sophistication of analysis and sophistication of presentation aren’t always the same thing.”
Personally, as a football fan and LPA player, I found the article insightful and somewhat amusing. It seems like the best way to solve the issue of hidden talents and unclear links between data scientist and team performance would be to gather more data, specifically on the data scientists, comparing their backgrounds, approaches and tactics. The article mentions first headhunting agencies to be evolving in the field, who likely focus on exactly that (and will require data scientists to support them). The industry seems to be growing and football fans, knowingly or not, are paying for it.