Event details from: Video Analysis and Artificial Intelligence in sport analytics: conflict or collaboration?

Date
25 June 2024
Time
15:00 – 17:00
Room
Sweelinck Room
Gian Marco Campagnolo

Video Analysis and Artificial Intelligence in sport analytics: conflict or collaboration?

The distinction between knowledge traditions based on visual evidence and those that describe phenomena in numerical and statistical terms is common to most scientific practices, from medicine to the natural sciences. One fundamental question in the sociology of science is: what are the consequences of these traditions for knowledge production? What are the tensions and conflicts between them and how can they be combined?

With respect to the evaluation of performance in sport, these two traditions and their allegedly antagonistic visions are discussed in popular books and movies. Modes of old-fashioned and “human” judgment such as traditional punditry and the “eye-test” have been pitted against new forms of expertise resulting from the rise of "quants" in sport analysis.

With examples from football analytics, in this talk we discuss what the field of sport can tell us about the “differences in vision” resulting from images vs data. Presenting early results from the analysis of a dataset of over 1,000 football matches, we compare how three different modalities – live commentary, video highlights and game stats – evaluate and classify performance.

We conclude with recommendations about the need for multi-modality to ensure diversity in how algorithms are trained for AI applications in sport and beyond.

Programme

15:00

Cake and coffee on arrival

15:30

Welcome & introduction by Huub Dijstelbloem

15:40

Lecture by Gian Marco Campagnolo (University of Edinburgh) & Markus Stauff (University of Amsterdam)

16:40

Q&A