A Bayesian Quest for Finding a Unified Model for Predicting Volleyball Games
share
Unlike what happens for other popular sports such as football, basketball and baseball, modelling the final outcomes of volleyball has not been thoroughly addressed by the statistical and the data science community. This is mainly due to the complexity of the game itself since the game is played in two levels of outcomes: the sets and the points (within each set). The final winner is the team that reaches first the three sets. Hence, the total number of sets is a random variable which ranges from a minimum of three to a maximum of five. In a second level, in order to win a set, each team needs to reach first a prespecified number of points (usually 25). Nevertheless, the number of points required by the team winning a set also varies depending on whether there is a margin of two winning points or whether the teams are playing the fifth set or not. In order to account for all these peculiarities of the game, we propose a unified Bayesian two-level hierarchical model. Implementation of our model on Italian Superlega 2017/2018, shows that our model is successfully replicating the final ranking of the league and outperforms in terms of Deviance Information Criteria (DIC) other models.
READ FULL TEXT