With proposals currently being accepted for the 2018 OptaPro Analytics Forum, we speak with Johannes Harkins, data scientist at OptaPro and OptaPro Forum judge.
To submit a proposal to present at the event, please click here.
What jumps out at you when you see a proposal? Submissions will include references to innovation, application, method and feasibility of the proposed work. Are there any aspects that particularly resonate with you here?
I think the best approaches succinctly lay out a logical approach to an interesting or new problem and clearly elucidate the reasons why it’s a valuable area of research. This means a thoughtful consideration of which elements of football tactics as well mathematical techniques are relevant to the problem at hand. It also means being well versed in the existing research on the topic and understanding what the potential pitfalls might be in the approach and the process of getting there from the raw data.
You’re in a unique position as the only judge that isn’t working for a club? Does this impact the way you judge proposals, and if so how?
I think the other judges and I broadly prioritise the same primary criteria. If I bring anything unique it’s perhaps a closer knowledge of the available data and a broader approach to the topics, where the judges in club roles may be more experienced with and focused on how these concepts can be applied in practice.
Is there anything - from a personal perspective at least - that you’d be particularly interested to see analysed at the OptaPro Forum? I know set pieces and throw-in situations are examples of topics raised in the past that might be of interest…
I think at this stage the football analytics community is poised to move beyond some of the single or limited-event metrics we’ve focused on in the past. Many analysts now are devising or relying on frameworks which create possessions out of multiple events, which opens a world of new possibilities. I’m intrigued to see how people use this framework to key in on specific tactical elements as well as more holistic notions of player added value.
The 2018 Forum will offer presenters the chance to work with outputs from Opta expected goal, expected assist and sequences & possessions models. What would your advice be to those looking to integrate these models within their analysis? Have you had any thoughts on how you’d like to see these models implemented and applied? This can of course also be extended to tracking data.
Though many analysts construct their own models and frameworks for these measures, I hope that what this opportunity offers is a chance to maintain some momentum in the research conducted at the Forum. By allowing people to begin from some of the basic places that past work has left off, we offer them a chance to focus on new and interesting ways to leverage these metrics, as well as compare them to novel methods that may introduce new levels of nuance.
Looking at this event from the perspective of football clubs, what do you think they should be looking to take from presentations, and from the event itself? It can be challenging for presentations, particularly when showing new work, to be directly applicable to all in the room, so should clubs be perhaps altering their thinking in how they approach the event?
I think it rather depends on the ways that a staff member’s club currently uses data in their process. Some people who are on the cutting edge might be looking deeply into the specific techniques on display. There may be others who are being shown detailed mathematical exposition and wondering how to bring this into a day-to-day process that currently relies on more traditional methods. To these people, I’d offer the advice to focus on the fundamental questions posed by the presentations, and what football situations they apply to.
Hopefully all the presenters will make this perfectly clear, but that’s not always the case, especially with such a varied audience. By focusing on what kind of questions people are posing about football, a club staff member can begin to understand what types of problems can be approached with data, and perhaps start to formulate their own football questions in a form that thoughtful data analysis might offer interesting answers to.
Looking at football analytics at a broader level, where do you think the industry is excelling, where has most progress been made, and where can most progress be made?
I think that analytics as a discipline has been more broadly seen as an important part of a club’s process, which is a sign of good progress.
This has come with a growth in the broader interest at a fan and media level in the type of narratives that analytics can produce, which is also an important factor in adoption and understanding of football analytics. I think the place to look now is building on previous work and approaching problems with more of a comprehensive understanding of how media and clubs have consumed analytics in the past. I think there’s a lot to learn from that experience that can be brought to bear on selecting the next set of questions that analytics looks to answer and how we present these answers.
Some proposals at previous Forums have been more suited to being presented as interactive posters rather than on stage. What are your thoughts on some of the posters you’ve seen presented at the Forum?
I really like the posters because they present a stage for certain projects which may be smaller in scope or incorporate more visual elements that wouldn’t fit as well into the format of a full presentation. They also offer the opportunity for interactivity which people have leveraged effectively in the past.
Do you have any favourite talks from past Forum, and if so then why?
Dean Oliver’s keynote speech in 2016 was a great way to round off the first forum I attended in person. I think he brought some great perspective as a foundational figure in basketball analytics and offered some sage advice to the whole audience. I also liked the way Will Gürpinar-Morgan approached the idea of stylistic elements of team play in the 2017 forum. I’ve also enjoyed hearing Will Spearman discuss some of what he’s been able to do with tracking data, both in 2016 and 2017. I think it presents the larger community with a good target for what types of things we should be striving to get out of that dataset and the interconnection of it with event data.
Click here to find out how to submit a proposal to present at the 2018 OptaPro Analytics Forum