B0619
Title: Forced to play too many matches? A deep-learning assessment of crowded schedule
Authors: Marco Delogu - Università degli studi di Sassari (Italy) [presenting]
Juan Tena Horrillo - University of Liverpool University of Sassari (United Kingdom)
Stefano Cabras - University Carlos III of Madrid (Spain)
Abstract: Do important upcoming or recent scheduled tasks affect the current productivity of working teams? How is the impact (if any) modified according to team size or by external conditions faced by workers? We study this issue using association football data where team performance is clearly defined and publicly observed before and after completing different activities (football matches). UEFA Champions League (CL) games affect European domestic league matches in a quasi-random fashion. We estimate this effect using a deep learning model, a novel strategy in this context, that allows controlling for many interacting confounding factors without imposing an ad-hoc parametric specification. This approach is instrumental in estimating performance under what-if situations required in a causal analysis. We find that dispersion of attention and effort to different tournaments significantly worsens domestic performance before/after playing the CL match. However, the size of the impact is higher in the latter case. Our results also suggest that this distortion is higher for small teams and that, compared to home teams, away teams react more conservatively by increasing their probability of drawing. We discuss the relevance of our results for decision-makers.
