In this paper, we deal with the problem of determining when and where a bus should short-turn on a single bi-directional line in real-time. We formulate a decision rule for when to short-turn among candidate short-turning locations that is based on the objective of minimizing total generalized passenger travel cost including waiting times and forced transfer. Computational results and analysis are provided via a simulation study in BusMezzo, a dynamic, agent-based transit operations and assignment model that represents both vehicle as well as passenger progression. The simulation framework allows us to evaluate the resulting trade-off between passenger costs and transit performance that occur when a decision to short-turn is made. The proposed short-turning strategy is applied to a real-world high-frequency transit line in Stockholm, Sweden.