During the past two years, eighteen different practitioners have been directly involved in researching and presenting key findings of proposals selected for presentation at Stats Perform’s annual Opta Forum.
Two of these presenters, Soumyajit Bose & Aditya Kothari, have subsequently gone on to work in the industry for Stats Perform, in the roles of AI Scientist and Data Scientist respectively, working alongside the company’s Chief Scientist Patrick Lucey.
Both of their projects focused on analysing defensive aspects of the game. Earlier this year Soumyajit presented alongside his co-collaborator, Abhishek A. Mishra, who together outlined ways of measuring and modelling the defensive efficiency of central defenders in the English EFL Championship, using only Opta event data.
A year earlier, Aditya introduced a physics-based measurement of each player’s defensive contributions on the pitch, using a model trained on tracking data to quantify each contribution.
Due to travel restrictions, both Aditya and Soumyajit had to present their findings remotely to delegates. Now, as 2023’s submission deadline approaches, Stats Perform caught up with both of them for a Q&A to discuss their experiences of presenting at the event and ask if they had any advice for practitioners interested in submitting a proposal to share their findings, if selected, in London next year.
When did you first become aware of the Forum?
Soumyajit Bose: I learned about the Forum mainly through the internet. This included Twitter, where the online football community share a lot of their work, as well as the football blog dedicated to FC Barcelona fans – BarcaBlaugranes. The people behind the blog are a really knowledgeable group, who share resources and tweets to make people aware of developments in the football research community. I also saw a lot of talks, from previous Forums, via Stats Perform’s YouTube and Vimeo pages.
Aditya Kothari: Possibly 2014, when I left a comment on the online video of a Forum presentation delivered by Colm McMullan. In fact, that may have been the start of my interest in football data science. I then started getting active in the football analytics community around 2018, which is when I found out more about the event, and the opportunities that were available for people to present.
When you submitted a proposal for the first time, what were your main motivations for doing so?
SB: I started fairly late in football analytics – it wasn’t until around 2017 that I first came to know about the fact there were people analysing the game through the lens of data.
Over the course of the next couple of years, I spent time revisiting old work and creating some new basic stuff, getting into discussions with like-minded people, and getting feedback and proposals for further work. This was a hugely rewarding process for me and the Forum provided a place to continue that, but in front of a wider group of more experienced and knowledgeable people.
The chance to directly work with a club analyst and also present work, if selected, to more people from the world of football was certainly very attractive. Also, by that point in my life, I had decided to switch fields from academia to football, and this was the perfect place to showcase the analytical skills I had to people who were professionally involved within the field of sports analytics.
AK: I thought I had an interesting idea to solve a problem that didn’t have many people attempting to work on it – measuring defence. I had written plenty of blog posts and tweets on various other football analytics topics, but this idea seemed like a bigger deal in terms of its novelty and the problem it was solving and I wanted to present this to a bigger audience and share it with people in the industry in the hope that someone might find it useful.
For each presentation you worked closely with an industry mentor. What did you learn most from them as they helped you build out your project?
SB: Jay Socik from Luton Town was our mentor. Both my teammate, Abhishek A. Mishra, and I already knew Jay, so it was a really smooth experience.
The biggest takeaway for me was that Jay had a great knowledge of the games’ nitty gritty and a vast amount of experience. He could think about scenarios where our analysis attempts would not give satisfactory answers. He compelled us to think, suggested improvements and encouraged us to showcase our analysis in ways that would appeal to the audience at the Forum. His thought process and eye for analysis is what I would like to achieve one day.
AK: I worked with Dafydd Steele from Liverpool FC and Thomas Seidl and Jonathan Whitmore from Stats Perform and they were a huge help in refining my presentation.
Their guidance on how to structure my talk, how to explain a complicated concept in a concise way, and what to highlight was immensely helpful. I could pack in a lot of relevant and interesting content for the various groups of people in the audience within the time I had.
What was the experience like of presenting your projects to an audience of club delegates, albeit in a remote setting?
SB: It was daunting but a proud moment as well. We got questions from club delegates, including from Tom Worville of RB Leipzig. It would have been even better if we could have been present physically and had longer chats, but unfortunately that was not feasible.
AK: Of course it’s a little more challenging to do presentations remotely but I still got quite a bit of engagement from people who saw the talk, met a lot of new people during and after the event, and got a lot of interest in my presentation after the event too. Except for the ease of doing a talk in person, I’d say the rest of the experience was as good as being there in person.
What response did you get from your presentation? Did anyone subsequently contact you to find out more about your project?
SB: There was some interest, mainly from people who were fellow hobbyists like me at that time. The overall response was pretty positive, I would say.
AK: I did have a lot of interest, maybe because of the topic as well. The Friends of Tracking group had me on their YouTube channel for a chat, and I also talked about my model on John Muller’s newsletter, Space Space Space. I also appeared on Analytics FC’s blog. I’m not aware of anyone in the industry using my concept, but if they are and if they are reading this then I’d request them to please let me know because that would make me very happy.
Finally, if you were going to give one bit of advice to anyone considering a submission for the 2023 Opta Forum, what would it be?
SB: Irrespective of whether a submission gets selected or not, you should be submitting a proposal. There are no downsides, and the upside is that you get to present at the Forum in front of industry and club delegates, and maybe get to talk to staff at your favourite club.
I see people bubbling with smart ideas on Twitter – just go ahead and put a submission in. Remember, you don’t have to create a production-ready model – you need to create a good enough proof-of-concept. You’ve already done part of the hard work by having these smart ideas. Now code something up, create a small demo and write-up, and submit.
AK: My advice is straightforward – the best ideas are usually simple and robust. Keep that in the forefront of your mind as you prepare your proposal.
The deadline for Opta Forum submissions is Tuesday 20th December.
Click here for full submission information, including details of each proposal category and data samples available.
