Prediction models for e-Sport matches in League of Legends

The presented bachelor thesis aims to explore the application of predictive models based on regression analysis for the League of Legends European Championship (LEC) from the game League of Legends (LoL). Methods of comparison, observation, data analysis, and document analysis were used. Data from domains managing historical match statistics were used to create prediction models. The tool which was used to process data and create all models was the programming language Python with library extensions for data analysis. All steps were done in Jupyter environment. The first task of the thesis focuses on modeling a logistic regression with minimal accuracy of 50 %. Subsequently, two additional algorithms for decision tree and random forest are created. All models are compared with the goal of finding the best option for predicting esports matches. Metrics of accuracy, precision, recall, and F1 score are used for evaluation. The models created provide a suitable basis for future developments of more advanced solutions for neural networks and real-time predictions.