Abstract

Sports analytics, with its dynamic and unpredictable nature, has gained significant attention due to its ability to improve decision-making and performance measurement. This study mainly focuses on predictive analytics in the context of the Indian Premier League (IPL), an annual T20 cricket league held in India. Various factors could influence the outcome of a cricket match, and hence it becomes challenging for analysts to identify patterns within the vast volume of data available. The proposed framework analyzes the most significant factors that have a significant impact on match outcomes and uses the factors to predict final cumulative scores of teams in the IPL after 20 overs. Deep learning techniques, such as Feedforward Neural Networks (FNN), Multilayer Perceptrons (MLP), Long Short-Term Memory (LSTM), and Recurrent Neural Networks (RNN), were applied on a dataset over ten years of ball-by-ball data collected from reliable sources such as ESPN and Cricksheet, covering 10 IPL seasons. The model was evaluated in terms of parameters such as Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R² values. Among the deep learning models, LSTM demonstrated the highest prediction accuracy, achieving an 85% success rate. The findings highlight the advantage of deep learning techniques in IPL score prediction with valuable insight for strategic planning and decision-making in cricket.

Authors

Ashish Raghav¹, Neeta Singh², Naresh Kumar³
Gautam Buddha University, India^1,2, University of Nizwa, Sultanate of Oman³

Keywords

Sports Analytics, Feedforward Neural Networks, Long Short-Term Memory, Dynamic Nature