A live cricket match and run prediction mechanism
Score prediction in cricket is a useful tool for teams, broadcasters, and other stakeholders in the game to make educated decisions and increase their engagement with the sport. Many authors have created a multitude of models to analyze the game, each seeking to predict and quantify various elements of matches, including runs scored, wickets lost, and overall winner predictions. Researchers have used both machine learning and deep learning models, testing for performance, accuracy, errors (MAE, MAPE, RMSE, etc), and execution time, with deep learning models generally having better results in comparison. The original dataset consists of ball-by-ball data for all ODI’s held between 2005-2020. After extensive pre-processing and feature elimination, an adapted version of the original dataset was achieved. Using this dataset, machine learning models (logistic regression, random forest classifier, and gradient boosting). In addition, to supplement consequent over-score predictions made by the mentioned models, two separate models, using logistic regression and decision trees, are used to predict the overall match winner. Drawing a comparison between each model’s performance and accuracy, the results of this research yield the optimum model. This work proposes a novel approach to the standard score or overall winner prediction.