Predictive Modelling of Player Performance in the Indian Super League Using Publicly Available Match Data: A Machine Learning Approach

Abstract

The global football leagues have widely adopted these techniques; Indian football is gradually catching up. In this paper, we present a machine learning-based approach specifically tailored for the Indian Super League (ISL) to predict individual player performance using publicly available match data. The collected and curated event-level data from over 300 ISL matches, focusing on features such as passes completed, tackles made, shots on target, and minutes played. From this data, we developed two predictive models: a regression model to estimate expected goals (xG) for each player, and a classification model to predict the likelihood of a player being named 'Man of the Match'. These models were trained and tested using standard machine learning techniques including Random Forest and Logistic Regression, and achieved encouraging accuracy and consistency. The results highlight that even with limited but structured data; it is possible to uncover meaningful insights into player contributions. This work serves as a step toward bridging the gap between traditional sports analysis and modern data-driven methods in Indian football. This study approach is scalable, accessible, and adaptable for teams, coaches, and analysts aiming to adopt a more objective and data-informed strategy.