Abstract
Today's competition between the professional eSports teams is so strong that in-depth analysis of players' performance is literally crucial for creating a powerful team. There are two main approaches to such an estimation: obtaining features and metrics directly from the in-game data or collecting detailed information about the player including data on his/her physical training. While the correlation between the player's skill and in-game data has already been covered in many papers, there are very few works related to analysis of eSports athlete's skill through his/her physical behavior. We propose the smart chair platform which is to collect data on the person's behavior on the chair using an integrated accelerometer, a gyroscope and a magnetometer. We extract the important game events to define the players' physical reactions to them. The obtained data are used for training machine learning models in order to distinguish between the low-skilled and high-skilled players. We extract and figure out the key features during the game and discuss the results.
Introduction
In the hyper-competitive world of eSports, understanding what makes a professional player successful is vital for scouting and coaching. Traditionally, performance analysis relies on two primary pillars: in-game statistics (such as Kill/Death ratios) or hardware logs (keyboard and mouse telemetry). However, these methods often ignore the "physicality" of the athlete. While it is well-understood that physical state affects performance, there is limited research on how a player’s body language and chair-bound movement correlate with skill level.
This research explores whether a player's physical "tells"—captured via a sensor-laden chair—can serve as a proxy for their expertise and emotional regulation during high-stakes gameplay. By analyzing how players move, lean, and react to specific in-game events, we can develop a more holistic view of the eSports athlete.
Methodology and System Design
To capture physical behavior, we developed a Smart Chair Sensing Platform. The core of the system is a hardware unit rigidly fixed to the underside of a standard gaming chair. This unit consists of a Raspberry Pi 3 connected to an MPU 9250, a 9-axis Motion Processing Unit containing an accelerometer, gyroscope, and magnetometer.
Data was sampled at 100Hz and transmitted via WiFi to a server for processing. The study involved 19 participants: 9 professional athletes from the "Monolith" pro team and 10 amateur players. Participants played the "Retake" modification of Counter-Strike: Global Offensive (CS:GO), a high-intensity mode focused on bomb-site defense and offense.
Data Analysis and Feature Engineering
The raw sensor data was synchronized with game logs to identify specific events such as kills, deaths, and shootouts. We extracted features like "intensity of movement" (using standard deviation in a 1-second sliding window) and "event-based reactions" (measuring movement within one second of a game event).
As shown in the sensor logs, specific game events like deaths (marked in red) often correlate with significant spikes in physical movement. Interestingly, the researchers found that the general intensity of movement—constant micro-adjustments and "wiggling"—was a stronger indicator of skill than reactions to individual successes like making a "frag."
Key Results
The study successfully demonstrated that physical behavior alone can predict player skill with high reliability. Using machine learning algorithms to classify the 171 labeled timeseries segments, we achieved the following:
- Classification Accuracy: The Support Vector Machine (SVM) achieved 78% accuracy in distinguishing professionals from amateurs.
- Model Robustness: The Random Forest model achieved a 0.88 ROC AUC score, indicating a strong ability to discriminate between skill levels.
- Top Features: The most predictive features were the median intensity of gyroscope data (representing wiggling or spinning) and movement patterns during active shootouts.
One of the most significant findings was that professional players differ most from amateurs in how they handle "tough" situations, such as being killed, rather than how they react to success. High-skilled players also exhibited more intense baseline micro-movements throughout the session.
Conclusion and Future Work
This research proves that a "smart chair" can effectively serve as a non-intrusive tool for eSports analytics. While the current study is limited by a small sample size and a single game title (CS:GO), it opens the door for real-time monitoring of athlete stress and fatigue. Future iterations could utilize Hidden Markov Models for real-time state detection, providing coaches with live data on a player's physical and emotional state during tournaments. Expanding the dataset to include other genres, such as MOBAs or RTS games, will further validate the universality of these physical signatures of expertise.
Summary updated