Influence of outliers in MOBA games winner prediction
Justyna Tyran, Łukasz Chomątek · Procedia Computer Science · 2021
This work regards about predicting the result of a match in Multiplayer Online Battle Arena team games. This type of games is characterized by very complex mechanics, which makes it difficult to predict the final result of the game based on its course. This prediction is based on the achievements of the teams participating in the game - usually complete data on the course of the match is available. Until now, the work in this field has not focused on capturing atypical games from the data set. For example -a team loses, although during the match its achievements exceeded that of the opposing team. The aim of this study is to examine the influence of atypical games on the quality of prediction process. The conducted research has shown that atypical games have a negative impact on the effectiveness of winner prognosis, and their removal demonstrates a beneficial effect on the prediction quality. The result of the work provokes further research in order to increase the effectiveness of predictions for matches with an unusual course.