This study introduces an innovative virtual reality (VR) training system for novice fencers, aimed at improving their offensive actions, distance perception, and overall combat performance. The system is divided into three modes: offensive actions training mode, distance perception training mode, and replay review mode. These modes work together to improve users’ precision and reaction time during combat. Additionally, we developed an action analysis algorithm that uses body skeleton data captured by the VR headset to recognize user actions and calculate their similarity to expert actions, allowing users to track their progress throughout the training. A pilot study with five participants preliminarily validated that the system offers an effective training process for beginner fencers. For a visual demonstration, see the video at: https://youtu.be/IhaavIannJo?si=O-ZLQs0iqhaQHJ9x .