Dynamic rupture inversion is a powerful tool for learning why and how faults fail, but much more work has been done in developing inversion methods than evaluating how well these methods work. This study examines how well a nonlinear rupture inversion method recovers a set of known dynamic rupture parameters on a synthetic fault based on the 2000 western Tottori, Japan earthquake (Mw 6.6). Rupture evolution on the fault is governed by a slip-weakening friction law. A direct-search method known as the neighborhood algorithm (Sambridge, 1999) is used to find optimal values of both the initial stress distribution and the slip-weakening distance on the fault, based on misfit values between known and predicted strong-motion displacement records. The yield stress and frictional sliding stress on the fault are held constant. A statistical assessment of the results shows that, for this test case, the inversion succeeds in locating all parameters to within ±14% of their true values. With the model configuration used in this study, the parameters located in the central rupture area are better resolved than the parameters located at the sides and bottom of the fault. In addition, a positive linear correlation between the mean initial stress and the slip-weakening distance is identified. The investigation confirms that dynamic rupture inversion is useful for determining rupture parameters on the fault, but that intrinsic trade-offs and poor resolution of some parameters limit the amount of information that can be unambiguously inferred from the results. In addition, this study demonstrates that using a statistical approach to assess nonlinear inversion results shows how sensitive the misfit measure is to the various parameters, and allows a level of confidence to be attached to the output parameter values.