Abstract
Counter-Strike: Global Offensive (currently Counter-Strike 2) is a highly influential first-person shooter game that has captured the attention of many players and spectators. Providing accurate and timely information regarding in-game progress is essential for enhancing the spectator experience and for a deep understanding of player strategies. In this study, we focus on predicting the winning team of a round by frame information from a particular moment within the round. Additionally, we investigate the key feature values contributing to this prediction. We developed a range of features, such as player health points, equipment, and spatial distances between players, and utilized the XGBoost to predict the winning team. Our results demonstrate that factors like the difference in the number of surviving players of both teams, the ammo of specific weapons, and the proximity between players and the bomb play a crucial role in determining the outcome. Moreover, we observe that the significance of these features varies depending on the map.
https://ror.org/05aevyc10 
