A new in-form and role-based Deep Player Performance Index for player evaluation in T20 Cricket

Sports analytics has benefited immensely from the growth and popularity of Machine Learning algorithms. Machine Learning and Data Mining advances have enabled sports analysts to evaluate a player’s performance more effectively. A review of existing literature on player performance evaluation methods shows the need to develop a new performance evaluation index for Twenty20 (T20) Cricket. We propose a Deep Player Performance Index (DPPI) to evaluate a T20 Cricket player based on batting and bowling strengths. DPPI captures a player’s current form and role in the team. DPPI serves a dual purpose. First, it enables sports fans and researchers to compare players playing a similar role in different teams. Second, the aggregated DPPI values of players playing at different positions in a team give the approximate team strength. To build DPPI, we first modify the existing Fédération Internationale de Football Association (FIFA) player performance evaluation guidelines. We then use the modified guidelines in the context of T20 Cricket. We propose DPPI based on K-Means clustering and Random Forest algorithm and compare our results with the existing player performance evaluation indexes for the Indian Premier League (IPL) 2019 season. Our empirical results show that DPPI captures a player’s batting and bowling strength better than other indexes. Thus, DPPI serves as a helpful index for fantasy Cricket users, Cricket fans, coaches, and managers to gain better insights into a player’s performance.