Hugo Rios-Neto's recap on FAME '23.
It’s been 10 days since Football Analytics: Modeling and Experience 2023 (FAME ‘23), and I finally got a chance to write about this wonderful event. As one of the main organizers, my expectations and uneasiness were quite high during the week of the event. Even though everything had already been arranged in advance: the recorded virtual talks, the recording of the event, a photographer, banners, the research competition, you name it, you only hope it all comes together nicely on the day of the event. And it did, beautifully, like we had all planned.
First of all, I would like to thank Wagner Meira and Adriano César Pereira, the two professors of the Computer Science Department (CSD) at UFMG who supervise our Sports Analytics Lab (SALab), whose research, teaching, and organization skills have been crucial to our lab growing so much in a year in a half. Also, I would like to thank all SALab members, who really stepped up in the final two months prior to FAME ‘23 and that, without their help, none of what happened would have been possible. Also, I would like to thank Gemini for sponsoring FAME! Without the extra budget, it would not have been possible to bring over so many talented speakers, who elevated the event’s quality so much. Personally, thank you, Jake, for coming to Belo Horizonte in person and taking part in this. I am sure SALab members look up to Gemini even more now that they know you and that we are only starting to nurture this talent pathway.
Starting off with some numbers, it’s clear we grew quite a bit from the first to this second edition. Last year, we had 250 people register and 175 attend. This year, 355 registered and 240 attended. Tickets were R$5.00, approximately a dollar in today’s exchange, just so that people didn’t register without having a clear desire to attend, with all the money being donated to a charity. In both editions, about 70% of the people registered attended. In terms of number of speakers and duration, the difference is significant as well. FAME ‘22 was a half-day event and had 12 speakers in total. This year, FAME was a full-day event and had 26 (!) total speakers. With so many interesting talks and panels, I’m just glad everything was recorded and that it will be made available soon.
FAME ‘23 kicked off with a presentation by Rodrigo Picchioni, titled “Enhancement of Football Operations Through Evidence-Based Practices”. It elaborated on the abstract term of football strategy, trying to define it more concretely. The presentation divided the concept of strategy into business and gameplay, aiming for success both on and off the field.
After that, there was a panel on the Quantitative Approaches in Sports Science with Monteiro, physiologist at Ceará, Túlio Horta, COO at Volts Sports Science and Health and Performance manager, and Adriano César Pereira, professor at CSD and moderator. The panel was able to showcase how data can play a crucial role in aiding decision-making in sports science. Despite both panelists having different demands in each of their work routines, both made use of data to support their prescriptions to athletes in how they should train and recover .
Subsequently, there were two quick presentations on papers we published at the lab this year. First, Leo M. Sá-Freire presented his paper “Field Depth Matters: Comparing the Valuation of Passes in Football,” which examines the valuation of football passes and their impact on the game. The analysis supports the hypothesis that the event-data-based VAEP is limited in distinguishing good passes in the initial third of the field, suggesting a spatial context limitation of the data format. The second paper, “Characterization of Football Strategies Based on the Frequency, Importance, and Effectiveness of Plays,” offers a novel approach to analyzing football strategies, building on top of frameworks like VAEP and SoccerMix.
Following, Alexander Schram, from PFF, SALab’s official data provider, showcased new ways to explore the game using their tracking data. Since starting to receive PFF’s data, we are now finishing structuring ourselves to be able to use the tracking data at scale and produce interesting research and content.
Our talk, “Building an Analytics Department with Gemini” began by acknowledging the difficulty in making data-driven decisions, emphasizing the communication barriers between data practitioners and less technical stakeholders, and the limited resources most teams face compared to larger organizations. The presentation introduced our platform, which is designed to automate foundational infrastructure and offer advanced no-code machine learning capabilities. This enables staff, including those with less technical expertise, to make informed decisions quickly. Gemini aims to replace ad hoc processes with streamlined operations, facilitating faster and more impactful use of data.
The final talk before lunch was the panel called “The Role of a Data Scientist in Football”. Composed of four panelists with diverse backgrounds and functions in the departments of the clubs they work, the panel was able to depict the multi-dimensionality of being a data scientist at a club. While in some clubs, the area in which analytics has the most impact is scouting, for others, it is match analysis, for instance. Another interesting topic that was discussed is how the club is structured, in terms of ownership, affects the types of demands.
After lunch, we came back with a virtual Q&A with Ravi Ramineni, co-founder and CEO of src-ftbl, hosted by our lab’s Ricardo Furbino, who is a data science intern at src-ftbl. Ricardo asked some really insight questions to Ravi about his trajectory and all things analytics. Following, a panel titled “Data Science as Support to Match Analysis and Scouting” had panelists comprising data scientists, analysts, and scouting experts, who explored how data analytics is transforming match analysis and player scouting. All panelists work for clubs or consultancy firms, so it was really interesting to note the similarities and differences they pointed out. They discussed the integration of advanced statistical models, and data visualization tools to analyze player performance and opponent strategies. Special emphasis was placed on how data-driven insights are aiding scouts and coaches in identifying talent and preparing for matches more effectively. The panelists also highlighted the challenges and opportunities in balancing traditional scouting methods with data-driven approaches, emphasizing the need for a synergy between empirical observation and statistical analysis.
After that, we had an hour-long coffee break where researchers were able to present their research posters to other attendees. The research competition was a notable addition to FAME ‘23. In it, we had 22 research posters accepted. Following the break, the two best submissions presented their findings on stage. Additionally, we had a presentation from a Physical Therapy professor at UFMG, Sérgio Fonseca, on the complexity of injuries in sports, a virtual talk from Caio Brighenti, Manager of Football Information at the Detroit Lions, on data science in the NFL, and a final panel that discussed the similarities, differences and collaboration opportunities between academia and industry.
Truly, couldn’t be happier of having helped put together such an incredible event, that was able to highlight the work of such amazing people and continue building an analytics culture in Brazil. Now it’s time to start planning for an even bigger FAME ‘24! Hope to see you then!