As eSports continue to gain popularity and become a professional industry, the need for effective training programs and accurate player evaluations becomes crucial. This is where the study presented in this article comes in, offering an efficient method to identify and classify the features that distinguish high-skilled players.
Importance of Identifying Distinguishing Features
Understanding the qualities and attributes that set elite players apart is invaluable for eSports coaches. By identifying these features, coaches can design targeted training programs aimed at improving and developing these specific skills. This specialized training approach is likely to result in more effective skill development and ultimately higher performance levels.
Additionally, game developers can benefit from this research by designing games that automatically adapt to players’ expertise. By understanding the distinguishing features that contribute to high-skilled play, game developers can create games that offer appropriate challenges and rewards, enhancing the overall gaming experience.
Challenges in Identifying Distinguishing Features
Previous methods of classifying eSports players using sensor data combined with machine learning have been effective, but they fail to provide sufficient information about the features that differentiate high-skilled players. This limitation hinders the development of targeted training programs and accurate player evaluations.
An Effective Method: Extracting Features and Classifying Players
The method proposed in this study addresses these challenges by introducing a novel approach. First, a time window is applied to extract relevant sensor data, such as heart rate and hand activities, before and after game events in the League of Legends game. This enables the capture of data during crucial moments of gameplay.
The extracted data is then utilized to calculate connectivity features between sensors using symbolic transfer entropy. This analysis provides insight into the harmony and coordination between different body parts, which can be indicative of high skill levels. The most relevant features are then selected using the newly developed consensus nested cross-validation method, ensuring a robust and accurate feature selection process.
Classification Results and Future Applications
The classification results obtained using this method are impressive, with an accuracy of 90.1%. This demonstrates the effectiveness of the proposed approach in identifying and classifying players’ skill levels accurately. Connectivity features between players’ gaze positions and keyboard, mouse, and hand activities were found to be the most distinguishing features in the classification process.
This research has the potential to revolutionize training programs not only in the eSports industry but also in traditional sports. By applying similar methods to sportspeople data, coaches and trainers can enhance their understanding of what sets elite athletes apart and develop more targeted training programs.
The combination of sensor data and machine learning techniques opens up exciting possibilities for both the eSports and sports industries. It allows for a data-driven and evidence-based approach to player evaluation and training, ultimately leading to improved performance and a more engaging gaming experience.