Evaluating Player Actions in Professional Counter Strike using Temporal Heterogeneous Graph Neural Networks

Research Paper will be posted in the coming weeks. Check back soon!
Download the
Full Paper Here
Authors

Szmida, Patrik Peter; Toka, Laszlo

Abstract

In recent years, electronic sports (esports) have gained popularity, extending the existing landscape of the sports industry. Counter Strike 2 (CS2), a first-person shooter team game, stands as one of the most prominent esports titles in 2024.In this esport, two teams face off within a match, taking turns as attackers (Terrorists - Ts) and defenders (Counter Terrorists - CTs). A match consists of2-minute rounds where Ts must plant a bomb at one of two bomb sites, while CTs must prevent it or defuse the bomb. The first team to win 13 rounds wins the match. With tournaments organized in front of large audiences and professional teams competing for substantial prize pools, the stakes of the professional scene are high. Despite these facts and the abundance of available data, only a few artificial intelligence-driven solutions have been explored so far regarding individual and team performance enhancement, and it has not yet gained much popularity in practical use.