23rd Conference of the International Federation of Operational Research Societies
Abstract Submission

620. Bayesian Models for Scouting and Multi-Season Squad Planning

Invited abstract in session FB-5: Bayesian Models for Football Analytics, cluster OR in Sports.

Friday, 11:00-12:30
Room: CE-204

Authors (first author is the speaker)

1. Pablo Galaz
University of Chile

Abstract

Professional football clubs seek to obtain the best possible performance in all competitions in which they participate; the Sport Manager must assemble, maintain and update a competitive squad at all times by negotiating contracts with incumbent and new players in the transfer market. A common strategy, adopted by the best clubs in the world, is to hire the best players within their reach, to improve the overall quality of their squad. Regrettably, this strategy is not feasible for every club, because of various constraints, e.g. budget and economic fairness rules. The problem of maintaining the best possible squad is complex, due to the many decisions regarding arrivals and departures of players required at the beginning of each season.

In this work, we formulate a framework for multi-season squad planning supporting the Sport Manager and the club in making decisions regarding the contracts of incumbent and new players, to improve the quality of the squad at all times while considering budget and other constraints. We estimate the performance of various squad configurations, and we use an analytical approach with event data from professional football to model match dynamics considering player-level interactions, to predict the development of a match.

The underlying model visualizes the development of a match as a discrete-time Markov Chain, whose transition probabilities depend on parameters, regarding the cognitive-perceptual and technical characteristics of the players. The approach enables measuring the influence of individual players on the collective performance of a team and can also be applied, for example, to guiding a team’s scouting process. We use a Bayesian inference approach to estimate the Markov Chain parameters, using data from 5 European leagues.
Using the above, we formulate the problem of defining a squad management policy using mixed-integer programming.

Keywords

Status: accepted

Back to the list of papers