Self-loop is a unique phenomenon observed in the daily operations of bike-sharing systems, characterized by bike returning to its original starting point after several trips within the bike mobility chain. The bike mobility chain concept involves forming new bike chains with a minimal fleet size. By understanding self-loop behavior, we can optimize fleet management and reduce operational costs. This study specifically investigates the self-loop behavior within the bike mobility chain while considering potential demand, using the case of the dockless bike-sharing system in Shanghai, China. An advanced multiply censored Tobit model is utilized to incorporate potential demand into origin–destination (O–D) data and reconstruct the bike mobility chain. The formation mechanisms of self-loop chains based on the land use and geographic location are analyzed. Our model achieved an R² of 0.871, significantly outperforming the baseline model. The results indicate that 76% of the bike chains can form self-loops within a 2-week period. Campus areas exhibit the highest self-loop rates, while suburban campuses can sustain operations with minimal or no scheduling required. This study not only reveals the back-and-forth behavior but also provides insights for scheduling and deployment strategies to enhance the environmental sustainability of bike-sharing systems.