Sports analytics: Designing a volleyball game analysis decision-support tool using big data

Abstract

From 2006-2012, George Mason University's (GMU) division I men's and women's volleyball teams were outplayed by their top competitors within their associated conference. Analysis of historic data showed that the GMU's men's and women's volleyball teams have a lower probability of scoring points on average of 0.21 and 0.05 respectively. The win/loss outcome is a function of the combinations of sequences of events caused by team's actions and coach's tactics. The data is so complex that no human can comprehensively conduct the analysis. A Computer-Aided Analysis Tool (CAAT) is needed to analyze the underlying trends contributing to the wins and losses as well as provide a meaningful recommendation to improve the overall team performance in a volleyball game. The CAAT determines the probability of each transition that can occur in a volleyball game, uses an Absorbing Markov Chain to evaluate how events influence the point scoring probability, and runs a Monte Carlo Simulation to analyze how random variations in transition probabilities, caused by extreme conditional scenarios can affect the team performance and end result of a game. Four design alternatives were identified through analysis of historic data and evaluated for improving team performance through specific skill improvement training: 1) Increasing aces; 2) Increasing kills; 3) Increasing blocks; 4) Decreasing errors. A utility analysis was conducted to determine the most effective design alternative to achieve the target level of performance. Based on the utility analysis, the GMU's women's and men's teams must focus on increasing their blocks. Out of 10 blocks, at least 9 should lead to a point for the men and 3 should lead to a point for the women in order to achieve the target level of performance.