Modely logistické regrese v oblasti esportových dat

Esport is a sport in a virtual world. Since the start of the 21st century, it has been growing amongst a new generation of people who grew up with computers, phones, and new technology. Esport offers classical sports, such as football, hockey, or rugby, as well as other types of computer, mobile, or console games. This thesis focuses on a computer game called CSGO. The topic is especially relevant because it is a new thing that continuously grows and evolves. The continuous growth affects, for example, betting shops, who are including esport matches in their day-to-day operations. This bachelor thesis aims to quantitatively analyze esport matches in-game Counter-Strike: Global Offensive (CSGO), predict winnings of both players and teams, and identify statistically significant predictors. The data used for this analysis are from kaggle.com and contain data between 2015 and 2020. The theoretical part deals with an introduction to esport history and an esport title called CSGO. The practical part focuses on data analysis in the programming language R. Predictions are based on a multivariate logistic regression model. Evaluation of predictions is done on a confusion matrix and resulting coming from it. The significance of predictors is then assessed using the Wald test. The result of this thesis are logistic models capable of predicting the results of a match based on player and team characteristics. Thanks to logistic models, significant predictors are identified, and their effect on the final result is assessed. These predictions help, for example, to identify a betting rate for betting shops.