"I got offered a dream job ... and said no."

George (not his real name) is a senior data scientist based in California. He has spent his career in tech, currently at a large AI tech company. He also has a long-standing interest in football analytics and has been doing work in that space on the side. About a year ago, he went through a full interview process for a football data scientist role at a US sports analytics company. He asked me not to name him or the company.

You weren't looking for a job when this came up. How did it happen?

A recruiter from the company messaged me on LinkedIn out of nowhere. There was nothing on my profile about football at that point. Nothing. So when he reached out, I don't think it was because I had any interest in soccer. He just saw my technical background and thought I could be a fit. But when I saw the role, it was a soccer data scientist title, fully remote, Bay Area based, and it kind of lit up my eyes a bit even though I wasn't looking. I thought, I should at least learn more about it.

What did you know about the company before you spoke to them?

Not much. I hadn't heard of them. They're not really a company the consumer market knows about. Their business is betting odds origination. They build the statistical models that betting companies use to set their prices. They started with the big American sports, NFL, basketball, baseball, and they'd been growing their portfolio from there. Football was one of the newer ones. They were hiring a couple of soccer data scientists specifically for it, partly with the World Cup in mind. For me that was kind of not ideal, because I'm not really into the betting side of things. There's an ethical side you have to think about. But I thought, it's still a soccer data scientist role. It could be interesting.

How did the process start?

First there was a call with the recruiter to understand my background. Then pretty quickly I was speaking with the head of sports analytics for the whole business. Someone who'd previously led the analytics department at both an NFL and an NHL team, so a very strong background. Forty-five minutes. And it wasn't writing code on a whiteboard or anything like that. He asked me hypotheticals. In baseball, if you were trying to figure out the efficiency of a right-handed pitcher against a left-handed batter, how would you approach that? What features would you use for a golf performance model? He was kind of ranging around different sports. Nothing football-specific.

Then what?

Then there was the take-home assignment. They gave me the 2015-16 Premier League StatsBomb event data for the whole season and asked me to build a predictive model for possession outcomes. What's the likely result of a given possession? Was it a successful pass? Was the ball lost? Something along those lines. The deliverable was a Python Jupyter notebook showing all my work. I had about a week. And it was pretty open-ended. They weren't being prescriptive about how I got there.

How long did you actually spend on it?

Probably more time than I needed to. But I went really deep and I just enjoyed doing that project. I built a really nice notebook, a bunch of visualisations, everything explained, all my features, the experiments, how this could feed into a calibrated XGBoost model, why I did that, how it could fit into betting odds. I was still working my normal job but I was spending all my nights on it. I was so into it. And I knew at that point I was going to do very well on that.

I tend not to give take-home assignments when I'm hiring. I think they can narrow the field in ways that aren't always fair. What's your view?

I see both sides. But for me, the reason I like them is that sometimes you can't fully show what you can do in just a conversation or on the spot. It's an opportunity to really show what I can do and that I'm passionate about it. And especially if you're an outsider to football, like I am, people might not give you a chance because you don't have experience at a professional team. But if you gave me the opportunity to show what I could do, I'm confident I could really impress you. A live technical interview is harder. You're having to speak out loud to strangers while you're coding and looking at someone else's screen. It's just not how you actually work.

What was the final interview like?

An hour with the manager of the soccer team and one of their data scientists. That was the most technical. They pulled up some Python code on screen and said: look at this, can you see any problems? That one was funny actually, because it felt weird on the spot. I know this stuff. But it took me a bit longer than I expected. There were also SQL questions. And then they asked me to build a statistical model for predicting how many goals a player would score. We worked through that and it got pretty technical. Then they pushed it: what if the model needs to be more dynamic? Because the whole point for them is that their system can generate odds on almost anything on the fly. I don't think I did quite as well as I should have done on that one. It's easy to get in your head in that environment.

What do you think they were actually looking for?

Overall I feel like their process was pretty similar to any other tech company. It wasn't anything super special about this being sports. The domain is sports, but really they didn't care too much about a deep understanding of football. Familiarity was nice to have. But what they really cared about is whether you can apply statistical modelling to the domain. That's it. And I think the same was true when I joined my current company. I didn't have experience in that industry either. If you're really technically strong and you can show you can adapt, sports is no different.

What did going through the process tell you about what the company was like?

It seemed fairly straightforward and organised. The majority of people there are technical. There's a centralised platform, sport-specific teams, a manager per team, and then the data scientists building the models. Fully remote, not a ton of meetings by the sounds of it, a pretty individual workflow. You own your model. And one thing they mentioned that I thought was interesting: you are responsible for your model once it's in production. If something goes wrong, if the model starts doing weird things, you need to fix it, because these betting companies are relying on it in real time. At a bigger company you'd have an ML ops team handling that. Here it comes back to you.

You got the offer. Why did you say no?

Two things mainly. The gambling side. There's the part of me that thinks, do I really want to be involved with that? And then compensation. My current role comes with stock compensation on top of base salary, like other big tech companies. The offer had a decent salary but no stock or anything like that. So when you look at total compensation, there was a pretty big gap. Between those two things, it wasn't the right move. This wasn't my dream job. My dream job is a soccer data scientist role at a professional club.

Any regrets?

I think for me the whole process was really cool and I got a lot out of just going through it. It kind of gave me confidence. I got offered a soccer data scientist role. I know I could do this if I really wanted to. And that was motivation for me to get more seriously involved over the last year and a half, which I have. So no. No regrets.

My take on this

The thing that stayed with me from this conversation was the contrast between how this company was approaching hiring and how most football clubs do it.

This was essentially how any tech company might hire. They put their own spin on it, but the shape was familiar. And they clearly valued fit as the initial bar, which I actually agree with. Fit and motivation are the most important things.

The contrast with football clubs is interesting, because fit matters there too, just in a completely different way. This company had a very high technical bar. The technical bar at most football clubs is often much lower, and what they compensate with is trying to set a trust bar very high. As an outsider, it's particularly difficult to get into a football team because you just don't have the network of people who can vouch for you.

If you've been through a process at a football club, an analytics company or a sports company and you'd like to talk about your experience anonymously, send me a mail on dominic@linear-b.co.uk