A Bayesian Classification Model for Predicting the Performance of All-Rounders in the Indian Premier League

The game of cricket got a new dimension, when the Indian Premier League (IPL), a competition of twenty over-a-side featuring eight teams named after various Indian cities/states started in 2008. The teams were franchisee driven and the players were selected via competitive bidding from a pool of available players. All-rounders i.e. players with the ability to both bat and bowl play a noteworthy part in cricket, whatever is the version of the game. The study measures the performance of all-rounders in Indian Premier Leagues (IPL) based on their strike rate and economy rate. The all-rounders are divided into four non-overlapping class viz. performer, batting all-rounder, bowling all-rounder and under performer. Stepwise multinomial logistic regression is used to determine the significant predictors responsible for such categorization. A Naive Bayesian classification model is developed that can use the significant predictors to forecast the class in which an incumbent all-rounder is expected to lie. The classifier is build based on the performance of all-rounders who participated in IPL-I and II, and the validity of the classifier is subsequently tested over the incumbent all-rounders of IPL-III. The classifier though moderately successful in predicting the appropriate class of the incumbent all-rounders in IPL III, is expected to perform better in future with increase in the size of training sample. This classification would be useful for the participating teams’ management while deciding about which all-rounder to be bided for and to what amount in the next addition of the league.