Smartboard: Visual Exploration of Team Tactics with LLM Agent

Area of Science:

• Sports Analytics
• Human-Computer Interaction
• Artificial Intelligence

Background:

• Team sports tactics are crucial for player interaction and game analysis.
• Current methods for visualizing sports tactics require complex reasoning to link actions to outcomes.
• There is a need for intuitive tools to understand and customize tactical plays.

Purpose of the Study:

• To develop a progressive approach for deeper understanding and on-demand customization of sports tactics.
• To create an extensible framework integrating large language models (LLMs) and visualizations for tactical analysis.
• To design an interactive system, Smartboard, for fine-grained tactical analysis and play design in basketball.

Main Methods:

• Collaborative development with basketball experts.
• Progressive sketching on a tactic board with agent-driven simulations.
• Integration of LLMs and visualizations for multimodal user-agent communication.
• Development of an agent-based interactive visualization system (Smartboard).

Main Results:

• Smartboard enables users to progressively understand tactical mechanics through step-by-step simulation.
• The system facilitates customization of tactics based on user-defined scenarios.
• Facet visualizations provide clear insights into simulated tactical actions.
• Case studies using real-world basketball data demonstrate the system's effectiveness.

Conclusions:

• Smartboard offers a structured process (setup, simulation, evolution) for iterative tactical exploration.
• The integration of LLMs and visualizations enhances user interaction and understanding of complex tactics.
• The system proves effective and useful for fine-grained tactical analysis and play design in basketball.