Football has changed a lot since 1966 and all that. In the 52 years of hurt (if you’re English) that have followed since, the game has evolved into a multi-million-dollar industry where even the smallest details are scrutinized by fans, players and coaches alike.
In the race for footballing supremacy, teams are using an increasing amount of data and information to help give them an edge over the competition. In 2009, when Manchester United goalkeeper Ben Foster credited watching videos on an iPad to a vital penalty shoot-out save, the idea was novel. Now, every goalkeeper does it. At this summer’s World Cup, England goalkeeper Jordan Pickford had notes written on his water bottle.
However, not every use of statistical data is this obvious to the fan watching in the stands or at home on TV. Clubs collect vast amounts of information about their own players and how they train, all of which can then be used by the manager and coaches to shape their approach for each match.
To learn more about this process and to find out what it’s like to be a data scientist at one of the world’s biggest football clubs, we spoke to Moscow Institute of Physics and Technology (MIPT) graduate Mikhail Zhilkin. Mikhail started working at Arsenal earlier this year, having previously worked for King – the makers of popular mobile games including Candy Crush. These two roles alone make his career path one of the more unusual you’re likely to find, so we were keen to find out how other STEM students could end up working in such unconventional roles.
Moving from King to one of the world’s leading football clubs is hardly a conventional career move. What drew you to the role at Arsenal?
I stumbled upon the position at Arsenal by accident. I’m subscribed to a regular analytics newsletter, and that’s how I got a list of vacant positions, which had Arsenal on it. Since I’m fond of soccer, especially English teams, I could not pass that up. I thought, in all likelihood, I would probably not get the job, because it was something quite specific and I had no relevant experience. But still, I submitted an application.
One of the interviews was about soccer: not so much about the numbers and data, more about understanding the game itself. Since I enjoyed playing soccer myself and watched TV shows with analysts discussing the games, I passed that interview. However, I’m aware that I’m the only person with no professional background in soccer and people still get surprised when they find out.
Were you a fan of Arsenal before taking this position?
I was a fan of Manchester United, but no one cares. Once you start working for a club, you gradually begin to support it. You can go to home games and cheer for the people you work with. A negative result casts a shadow over your work as well. You know how soccer is, lose a match and everyone is tempted to find someone to blame, when in fact it could be just about bad luck.
Did your experience with Candy Crush help at Arsenal? Were there any skills you carried over from your previous job to the new one?
Those were the most basic skills: knowing how to write a query to a database and perform statistical analysis using a programming language. The ability to work in Excel is also important. People may underestimate this. In my first year at King, I mostly worked in Excel, and I was doing fine. Today I get some of the data in spreadsheet format. If you can solve a problem in Excel, why not?
What does your job actually involve on a typical day?
One of the things that makes my job special is working next to the other people in charge of the physical training of the athletes. We go to the same rooms, put on the same clothes, change in the same room. I’m present during the daily briefing, even though the focus of the discussion is on injuries, fitness regimen, and the appropriate intensity of training. I’m not knowledgeable about this, but I’m involved in the process, so I go to these meetings.
I work in very close contact with these people and the athletes themselves. Much depends on the team’s schedule. Generally, once the season has started, they have one training session before lunch. I’m there to help with the simplest of things.
For example, during the session an athlete wears a special close-fitting jersey with a slot for a GPS transmitter, so we can track their movement, acceleration, deceleration. They also wear a device that tracks the heartbeat. Someone has to prepare all that stuff. Furthermore, you need to make sure everyone puts on their GPS transmitter and does not mix things up.
When people return from training to have lunch, you need to pull out each GPS transmitter from the sweaty jerseys. You get to see how your data is collected, so you can identify problems and possible inaccuracies right away. Say, if someone didn’t put on their transmitter or forgot to turn it on, you recognize the problem and find ways to interpolate the data somehow.
You also face issues such as: What if we didn’t collect the data on a player and we need to figure out at what intensity he’d been training, how can we gauge it, at least approximately? If you’re into soccer, this behind the scenes stuff is interesting. Plus, you get to work in really close contact with the athletes. Some people spend days outside the training ground just to catch one of the players driving out and ask for an autograph, while you get to take a sweaty jersey from him.
When assessing player fitness and performance, what are the key pieces of statistical data you’re looking at?
My team doesn’t assess the performance of the players or tactics, though data analysis could be employed for that. In soccer, unlike baseball, the objective indicators, such as the distance run or balls passed cannot be interpreted in a straightforward way.
If you count up the number of accurate passes made by a certain player, this wouldn’t give you a complete picture, because someone tends to pass the ball backward or sideways, for example, which doesn’t help the team to advance. Another player might attempt riskier long balls, most of which get intercepted, but if they don’t, this creates an opening and leaves the forward one on one with the goalkeeper. So, it’s much more subtle.
For now, our main objective is to capture a complete picture of how prepared every athlete is and try to optimize training intensity, because a soccer player is treading a fine line between undertraining and an increased risk of injury. You have to find the middle ground where he is alive and well but also in top condition for the next match. In soccer, winning the next match is always a priority.
You’ve worked in two industries where there’s a very clear metric for success, whether it’s the number of downloads a game has or the number of matches a football team win. How much connection do you personally feel to those successes?
So far, the contribution is minimal as I’m just starting out. I have no grand ambition, because in soccer, the result in many ways depends on chance.
Up until now, I’ve been trying to help with the little things: give people more information for making decisions, so they would be more accurate and objective. Perhaps a year or two from now we will be doing advanced stuff that would have more of an effect on decision-making. The decisions that have to do with training and preparing the athletes for a match are only one piece of the puzzle.
I wouldn’t say the things I do make a decisive impact, but we’re talking about a top club playing a professional sport and big money is involved. So even if you can nudge the balance slightly in favor of your team, this might well be enough to make your work worthwhile.
When you were a student at MIPT, how did you picture your future career?
Everything kind of happened by itself. I got into King almost by accident. I realized that analyzing data could be interesting for me, but I had no carefully thought-out plan. I tend to begin work in the hope that it would turn out interesting and try to obtain the necessary skills in the course of projects. I don’t see myself as a highly qualified specialist in any particular field.
What advice would you have for students reading this who are interested in science and data, but aren’t sure of the career paths into roles like the ones you’ve occupied?
My advice would be, first and foremost, that one should learn English, and that goes for almost anyone, regardless of which field you want to work in. The fact that English was taught at a decent level at MIPT was important, perhaps even more so than math or physics.
I’d also recommend trying out different things, different jobs, different projects. It took me a long time to realize that I liked to work with data. You probably will never know, unless you try. Expand your horizons, don’t be afraid to try something new.
