Recently, sport video analysis has attracted lots of researcher¡¦s attention because of its entertainment applications and potential commercial benefits. Sport video analysis aims to identify what trigged the excitement of audiences. Previous methods rely mainly on video decomposition using domain specific knowledge. The study and development of suitable and efficient techniques for sport video analysis have been conducted extensively over the last decade. However, several longstanding challenges, such as semantic gap and commercial detection are still waiting to be solved. In this work, we consider using semantic analysis to adjacent pitch scenes which we called ¡§gap length.¡¨ Difference kinds of baseball games show its specific distribution for gap length, which depicts the potential significance of each baseball game.