Published on: 22/12/2022 at 3:16pm - Last edited on: 23/12/2022 at 7:51pm

Seminar on data science and sport

On December 15th, the French statistical society (aka. SFdS - Société Française de Statistique) together with INSEP (Institut national du sport, de l'expertise et de la performance, the French national institute of sports) organized a seminar on the applications of data science in sports. Christophe Ley was the keynote speaker for this event and Florian Felice presented his PhD research in 180 seconds.
The event was organised in one of the greatest French institutions of sports, INSEP. This institute focuses on training profesionnal athletes with the objective to help them get to the gold medal in the next Olympic Games of 2024 in Paris.
The talks were meant to be for a large audience to promote the possible application of statistics and data science in the field of sports.

Data science and sport as a gold or silver combination

With the objective of illustrating, in a pedagogical way, different applications of data science and their impact on sport, Christophe Ley attributed medals (from gold to bronze) to existing fields of applications.
A gold medal could be attributed to analysis of strategies. As the volume of data available keeps on growing, and more sensors are used during competitions, analyzing strategies of players/teams becomes more and more common.
The silver medal could be attributed to the prediction of games scores. Several applications exist, with the most commonly known from bookmakers, but the maximum level of accuracy tend to remain quite low depending on the sport to predict.
Finally, the bronze medal could be represented by the field of injury prediction. Some applications already exist to help athletes avoid undesired gesture/activity that could lead to an injury but these applications quickly face an important problem: the lack of data. The current research activity and the increase of tools to collect athletes' data make this field particularly promising for the upcoming years.

Combination of statistics and machine learning with applications in sports

In the spirit of the competition "Three minute thesis", Florian Felice presented his PhD objectives in 180 seconds. The goal of his research is to combine the strengths of statistics (with provable mathematical properties) and machine learning (as highly performing predictive models) tools to take the best of both worlds to apply to multiple domains such as sports.
One can combine statistical features to estimate unobservable quantities (i.e. variables derived from distributions of available covariates) to augment the data set used to train a machine learning model. In the field of sports, this can help estimate parameters such as the strength of two opponent teams.

MIDAS as a key player in the field of sport analytics

With the MIDAS team being represented during the seminar, and in particular with Christophe Ley as the keynote speaker, this shows the place the team is taking in the research applied to sports. It highlights the current achievements and strengthens the research direction for the team.
During the seminar, both talks received several interesting questions and highly positive feedback. These may lead to more future collaborations between sport experts and MIDAS research group, so the team can continue to apply its statistical and machine learning expertise to help athletes in their search of medals.

Find more information