Award for Technology Transfer Perspectives at the PhD Workshop - Adrià Arbués at the MIT Sloan Sports Analytics Conference

(Text by Adrià ArbuésGPI research group - winner of the mVentures - The Collider Award for Technology Transfer Perspectives with his work “Identifying Basketball Plays from Sensor Data; Towards a Low-Cost Automatic Extraction of Advanced Statistics”)


Overwhelmed. This is definitely how I feel right now. Since I finished my bachelor degree, I thought that the world of data science applied to sport was tiny, and that I could be (kind of) a pioneer in this field by applying Computer Vision and Artificial Intelligence algorithms in basketball sequences… FALSE! I just came back from the MIT Sloan Sports Analytics Conference (Boston), where I might be the less nerd in the tech-sports field among the 3500 attendants, what an awful feeling! In this conference, experts (Paul Pierce, Sue Bird, Adam Silver…) explained the current sports trends in different panels, companies held meetings with students in their boards and showed their research as well as offering job opportunities, and a research paper competition was taking place too; all happening at the same time, complete madness.

Although I was not presenting any scientific paper, and my status-quo is not even close to the threshold to be considered an expert, I could have the chance of attending this Conference thanks to mVentures – The Collider Award for Technology Transfer Perspectives, obtained in the 6th Doctoral Student Workshop (2018). Having no presentation-pressure or whatsoever, I could attend to approximately 15 talks, and had time to chat with many companies to get insights on how to enrich my research plan. For the first time, I had the chance to talk with people that had to deal with the same programming issues that I am facing at the moment, and I obtained answers to geek questions such as “at which resolution must a video be in order to detect proper arm limbs in an occluded player?”.

Besides, I discovered that one of the two main tracking NBA companies (STATS) recently launched a product (AutoSTATS) that performs multi-tracking of players using pose information in broadcasting videos, which is the exact same thing I am currently working on (even using the same libraries); however, it is crystal clear that with their large amount of data, it is impossible to compete against their results. Although it was quite frustrating at the beginning, we are currently trying to find a way of collaboration not to overlap with each other. Being in touch with these companies, and being able to talk with the real pioneers (such as Patrick Lucey or Luke Bornn) is just priceless. A really PhD-changing experience.