Gaming Telemetry Analysis Application Architecture
- 1. Introduction In recent years, visualization of gaming telemetry data has become increasingly important, enabling players to make data-driven decisions based on performance and behavior analysis. Our first quarter goal was to develop real-time visualization tools that allow players and coaches to collect and analyze actual gameplay data. To achieve this, we developed two applications: 1. A desktop application for real-time 3D visualization 2. A web application for post-match analysis This presentation describes the overall technical architecture and challenges faced during development.
- 2. Project Overview Gaming Device Desktop Application (real-time visualization) Architecture Pyside6 + PyQt + OpenGL Primary functions - UDP Packet Processing - Data Logging + Cloud Upload - Real-time Visualization Web Application (post-analysis) Architecture Streamlit + Firebase Admin Services + Google Cloud Storage Primary functions - User authentication - Search specific data from Cloud Storage - Data Analysis Google Cloud Storage Telemetry UDP Packet Store JSON List & Load JSON files for post analysis Retrive JSON to compare with realtime data Data Flow
- 3. Effective Development Environment Desktop Application The project originated from the community's existing discovery of UDP packet accessibility from a gaming device. Building upon this known capability, we focused on creating a comprehensive visualization tool that players can analyze their own data. We selected PySide6 as our development framework, leveraging its robust features for real-time data processing and visualization. This choice enabled rapid development while maintaining high performance, all within a pure Python environment. Network Layer Decryption UDP Socket Binary Data Processor Application Layer Thread Manager User Event Signals Graphics Layer Store Data Real-time Visualization UI State Update System Architecture Processed Telemetry
- 4. 3D modeling visualization Our application transforms telemetry data into 3D motion using OpenGL through PyQtGraph, which which will be replaced by VisPy. The visualization engine implements quaternion-based rotations and matrix transformations for precise object movement while providing multiple camera control options. Desktop Application Physics Visualization Based on matrix calculations translate complex motion data into smooth visual transitions, with temperature-based color mapping. Camera controls Multiple viewing angles and position tracking enhance data interpretation through intuitive camera controls.
- 5. Thread Management Thread separation is crucial for our application since it requires to handle continuous data streams with simultaneous processing, visualization, and logging. This handling improved the stability of GUI rendering performance and avoid possible logging data loss. Packet Process Thread | Data Logging Thread Main Thread Desktop Application Start / Stop Signals User Events Signals
- 6. Quick Prototyping of Data Analysis Dashboard Web Application Our first quarter focused on getting ideas and insights from players by using real data, leading us to choose Streamlit for its simple backend development, deployment environment, and user authentication capabilities. The framework allowed rapid development, enabling multiple stakeholders to access and utilize the application with real data. Architecture Streamlit + Firebase Admin Services + Google Cloud Storage Primary functions - User authentication - Search specific data from Cloud Storage - Data Analysis
- 7. Data Manipulation & Flow Our application combines three powerful Python modules: Pandas, NumPy, and SciPy, each handling specific aspects of telemetry data processing and visualization by Plotly. Web Application Data Processing Plotly The processed data is visualized through interactive plots such as scatter plots, line charts, and path visualizations. Users can explore temporal data sequences, compare multiple datasets, and analyze performance patterns. Interactive Analysis Pandas + NumPy + SciPy Raw JSON data is transformed into dataframe and manipulated by Pandas, NumPy, and SciPy. This structured data includes position metrics, performance indicators, and identified key points through signal processing.
- 8. Advantages of using Object Storage for our JSON files Object Storage maintains data in its raw format, eliminating the need for data transformation. It is a suitable option for our project since the desktop application dumps telemetry data directly as JSON. The filename format follows username_identifier_date.json. Search functionality is implemented by utilizing Google Cloud Storage's list_blobs API with prefixes, where prefixes can be either username or username_identifier. This simple approach is sufficient for our current data volume and usage patterns. Prefix based Search Web Application Users can search by username and identifier, for instance, “alice” and “002”, in which two files will be listed in this example.
- 9. Challenges Performance Impact of Page Refresh Streamlit's automatic page refresh mechanism poses performance challenges in our application. When a user modifies any parameter, the entire page reloads instead of updating individual components. This full-page refresh affects user experience, especially when interacting with our multiple interactive visualizations. As a temporary measure, we reduced data points for visualization using pandas DataFrame slicing (df.iloc), though a more fundamental solution will be needed. Web Application From Self-Service to Guided Analysis We are shifting from providing self-service dashboards to delivering focused analysis reports with regular user meetings. While our initial plan was to develop comprehensive tools for users' self-analysis, feedback showed users preferred analysis of specific scenarios based on their individual needs. For the next quarter, rather than building an all-purpose dashboard, we're adopting a market-in approach focused on targeted analysis based on user feedback. This shift allows us to utilize existing tools like Jupyter for quick visualizations rather than developing a complete dashboard.