A Prototype System for Training Method Recommendations for Baseball Players Using ChatGPT

Abstract

This paper presents a prototype system that applies ChatGPT-4 to personalized baseball training support. The system generates training menus by analyzing six user inputs: height, weight, dominant hand, fielding position, strengths, and weaknesses. Unlike conventional coach-dependent guidance, it provides concrete and easy-to-follow suggestions such as “100 swings,” along with relevant keywords to enhance usability. An evaluation with five inexperienced and five experienced players assessed UI, response time, and training quality. Both groups rated UI and response time highly, while inexperienced players found the training suggestions useful. In contrast, experienced players reported insufficient specialization and practicality. These results demonstrate the potential of generative AI for sports training support, particularly in assisting beginners, while also highlighting the need for improvements in prompt design and domain-specific knowledge integration to meet the demands of advanced users.