The two-step stochastic Empirical Green’s Function (EGF) method does not rely on kinematic descriptions of source rupture processes, offering significant advantages in simulating ground motions for target earthquakes. In this study, we employed the two-step stochastic EGF method to simulate the ground motions of the 2021 Yangbi MS 6.5 earthquake by selecting observed re-cordings from nine small earthquakes in the Yangbi earthquake sequence as EGFs. We focused on investigating the impacts of EGF selection, rupture directivity of small or target earthquakes, and stress drop on the simulation results. The following conclusions were drawn: 1) EGF selection significantly influences simulation results. When different EGFs are selected, the standard deviation of simulated records for the target earthquake at the same station approximates the inter-event standard deviation of small earthquakes, reflecting the mapping of source effects from small earthquakes into simulated ground motions; 2) When observed recordings from a single small earthquake are used as EGFs, the uncertainty in simulated records reflects the variability in ground motions related to propagation paths. This demonstrates that the EGF method effectively captures path effects in simulated ground motions; 3) Stress drop and rupture directivity effects must be carefully considered in simulations. Deviations in stress drop cause systematic errors in simulated recordings, while rupture directivity significantly affects the spatial distribution characteristics of simulated ground motions.