Skip to Main Content

Key Takeaways From Sloan 2019

By: Stats Perform

The start of the year is always a hectic time for sports analytics events and therefore, my calendar.

From January through to March, there’s one OptaPro event per month, each in a different country. This is my version of the Saturday-Tuesday grind faced by many in the industry, and I applaud those who go through it without a fuss.

The beginnings of 2019 have been slightly different. MIT’s Sloan Sports Analytics Conference was a welcome side dish of multi-sports-analytics-mega-event, helping me to properly deconstruct and digest my main course of football analytics conferences.

Up until a few years ago, football was only a small part of the debate at Sloan. This year, however, we have seen multiple football-specific panels and talks, several posters on football and the winner of the research paper competition being focused on football. The beautiful game can no longer be said to be too fluid to analyse. Squeeze up a bit basketball, shuffle over baseball, there’s room for us at the table for major sports influenced by analytics.

Below are my thoughts on a couple of talks that I enjoyed, and three questions I have when comparing my experiences at Sloan to the current state of analytics in football.

Shaping the future of football

My first session was the OptaPro-sponsored ‘Shaping the Future of the Game’ panel. The discussion was wide-ranging, touching on topics such as the required characteristics of young players to be successful, which metrics can be used to predict whether a player will become a senior professional or not and the future impact of VR on helping players learn and improve.

The discussion around player development is one that impacts both football here in England (because of the small issue of Brexit and difficulty in obtaining working visas) and in the US (fewer teams are deriving value from the SuperDraft in MLS and more are investing in their own academies).

It’ll be interesting to look back in a few years’ time and see how processes have changed from a scouting perspective as the sample of youth data grows, and a new generation of coaches at the helm of the sport. With such a different pathway in football compared with traditional US sports, this is an area where football can shape its own analytics future.

While in the next five years, clubs will undoubtedly progress at different rates (just as we have seen over the past five), the key areas that will become the norm in the context of data and analytics informing player progress are likely to surround loan management and the provision of appropriate playing experiences. While this is taking place to a certain degree, this currently stops at confirming a team for said player, rather than exploring how a loan impacts their future development.

This will extend to the number of players that turn professional with their clubs, rather than being sold on or dropping out of the game. A club that understands what the modern footballer looks like now and in five or so year’s time in their given league will be able to tailor an academy programme to develop players with a skillset geared towards this goal.

Hunting for unicorns

Next, I attended the basketball analytics panel ‘Hunting for Unicorns’. With the panel tipping off with a discussion about what a ‘unicorn’ is (essentially a player who is a complete outlier in terms of skillset/profile). I immediately saw parallels to the types of discussions we have in football, both those involving mythical animals (“can you win without a superstar/unicorn?”) and those that don’t (“what would happen if you put former-great-player-x in the league today?”).

There is often the debate about how team x won title y, and it was because of their unique z playing style. We then see other teams replicate style z, and the styles don’t change until another team breaks out of their orbit and starts to zag while everyone else zigs.

As Paul Pierce noted on the panel, the NBA is a ‘copycat league’. While we don’t see team styles mirrored to the same extreme in football as in basketball, it was a good discussion on how to find an edge and how thinking (but not shooting!) outside of the box can bring gains – as long as you keep fighting inertia.

I feel that there’s a media narrative in our sport of how teams look to adopt the style of the team du jour (Leicester City win the league, possession is dead! Manchester City win the league, long live possession!) where the issue is a blindness to the fact that a game of football can be won in many different ways. The “Unicorn” panel all agreed that while the Milwaukee Bucks, Houston Rockets and Golden State Warriors put up similar numbers, they all play a very different, incomparable brand of basketball.

There was also a great piece of advice from Mike Zarren, Assistant GM at the Boston Celtics, about how an analyst should say nothing when they don’t know. Dean Oliver made a similar point in his keynote speech at the OptaPro Forum a couple of years ago, and I can certainly remember situations where I wish I’d subscribed to that line of thinking in the past.

I briefly mentioned the posters earlier, but I was really impressed with the two football ones on show – partly because both of them were built using only event data. To me, this is a great rebuttal to the idea that we’ve wrung every last drop of utility out of this dataset. Statsbomb’s Derrick Yam’s poster identified a framework for analysing goalkeeper performance, while SciSports’ quartet broke down playing performance under different facets of pressure (pressure in terms of game importance, not Roberto Firmino breathing down your neck).

Future challenges

Computer vision was a major theme throughout the conference this year, with plenty displaying outputs powered by data generated through this approach. The underlying message seemed to be that we need more granular information on body pose to answer questions posed by coaches. I feel that, while that may be true for the more analytically advanced sports it may be a case of too much too soon for football. While I don’t disagree with the sentiment that better-quality data can inform better quality analysis, it does open up several challenges for our segment of the industry going forward.

The first of these is considering how to best to get value from tracking data. To me, it seems that the real value in tracking data is to give us better denominators. For example, taking the total number of forward passes and being able to understand whether they are from situations where a player is under pressure, is breaking a line, or passing to a team-mate in space. Being able to enrich event data with qualifiers that analysts don’t have the capacity to track live or are too labour-intensive to track post-match is where most of the value is going to be in the short to medium term.

The second of these is to consider what the solution is to fill the skills gap between those analysts who can derive value from event data, and those who can do the same from tracking data. With event data, people such as Rob Carroll have given video analysts the tools to go ahead and produce meaningful insight from event data, just using Excel. This approach cannot be replicated with tracking data. In my mind, the leap in complexity between these datasets pushes the onus onto the builders of products, or it pressures teams to hire someone with the skills to use this data. While some companies displayed some truly innovative approaches to capturing data from video, the challenge will be scaling this up and providing the data into an easy-to-digest-and-analyse format.

Finally, I foresee there being a shortage of what I’d call ‘translators’ in the industry. There are probably only a handful of people right now who have a job like this, which involves designing and implementing a strategy around the use of data within a club and has a relevant tactical understanding of the game. This person is responsible for choosing what technology, data and products are used, the personnel who are hired and is adept in understanding the needs across the academy, recruitment and analysis departments. They’re not (always) doing the more technical work, but they have a complete understanding of it. As teams begin to hire more technical staff, having someone non-technical to pull it all together seems like a vital step to getting buy-in and actually feeding into decision making processes.

To those looking to break into the industry, I’d offer this as a must read. I feel that there’s a supply shortage in those people who fully understand how clubs work across analysis and recruitment, the product offerings available in the market and the mix of skills required for a successful data analyst or scientist at a club. While football analytics is a predominantly technical field, there’s definitely a need for those who can write the playbook, and leave the execution of the plays to others.