This study proposes a multi-level comparative framework for analyzing human and engine opening strategies in Chinese chess, integrating sequence similarity, structural similarity, and coverage metrics to comprehensively examine differences in strategic distribution, variation choices, and opening tendencies. The framework captures both complete move sequences and local node structures, while incorporating weighting and threshold adjustment mechanisms to flexibly balance strategic representativeness and coverage according to research objectives. The findings reveal the concentrated and robust nature of human play in classical variations, as well as the engine’s advantage in diversity and innovation, highlighting structural differences in their overlap on core strategies and divergence in unconventional variations. This analytical approach not only enhances human players’ understanding and adaptability toward engine strategies but also provides engine developers with insights for optimizing opening books and designing more diverse and challenging strategies, thereby advancing strategic learning and applications of artificial intelligence in game-playing systems.