Abstract:
|
In team sports such as basketball and hockey, the plus/minus metric is used to quantify each player's intangible value to the success of the team. In order to calculate this metric, one needs each person's time on the court/rink, and the team's points scored and allowed during this time. Volleyball features a high volume of scoring and player substitution, two qualities which make the plus/minus an interesting calculation. Despite this, its use is not wide spread in NCAA volleyball. A major reason for this is because the primary data collection instrument does not record the movement of the libero. In other words, it is often uncertain as to which players constitute the back court for a possession. To handle this incomplete data issue, we take a Bayesian approach, combining a multinomial regression model of court presence with a standard regression model to estimate the plus/minus metric. We measure the performance of this model by comparing its result against the gold standard video evidence for a certain Division I volleyball team for which back court presence is known for only 15% of plays.
|