Analyzing key factors influencing IPL cricket scores using explainability and multimodal data

In this study, we investigate data from the Indian Premier League (IPL) spanning from its inception in 2008 to the most recent 2024 season to identify and analyze key factors influencing cricket scores. Using the H2O AutoML framework, we develop a predictive model focused on identifying low first-innings scores, incorporating data on location, weather conditions, teams, and players, while distinguishing them from matches with par or high score. Explainable AI (XAI) tools are employed to quantify the influence of various match features on score predictions, ensuring transparency in the model’s decision-making process. To further enhance classification performance, we introduce pre-match pitch report descriptions generated by a Large Language Model (LLM). For a subset of matches, we leverage multimodal LLM capabilities to analyse pitch report videos, comparing their predictive value against textual descriptions. Our findings underscore the potential of AI and machine learning in sports analytics, specifically in predicting cricket scores based on pitch conditions and other influential factors. This research provides valuable insights for teams, coaches, fantasy sports enthusiasts, IPL administrators and analysts, helping to optimize strategies based on available pre-match information. As part of our work we are sharing a pitch report dataset, python source code for the predictive model with explainability, and a Most Valuable Player (MVP) implementation framework to enhance reproducibility and support further research in cricket analytics.