Live Cricket Predictions for Runs and Win Using Machine Learning

Runs prediction and win 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. This work used machine learning models to predict the runs scored by a team in the first innings and to predict the winner of a cricket match on a real-time basis. The Cricksheet dataset which contains ball-by-ball data for international matches played between 2005 and 2020 was used. For the runs predictor, the data was pre-processed to achieve the desired features required to predict the runs of a team at the end of the first innings. Similarly, for the win predictor, the dataset was pre-processed to obtain the required features for the second innings. The best-performing model for the runs predictor was XGBoost with an R2-Score of 0.99 and RMSE of 5.85. Meanwhile, Random Forest was the best-performing model for the win predictor with an accuracy, precision, recall, F'l-Score of 99.99%. However, further analysis of the results indicated that the Logistic Regression model was better at predicting the match-winner probability. A Graphical User Interface was developed for both win and runs prediction using the best-performing models.