The U. S. Weather Research Program Hurricane Landfall has sponsored the Joint Hurricane Testbed program to facilitate a transition of research toward operations at the Tropical Prediction Center/National Hurricane Center (NHC). This paper describes the results and findings of test of the revised Dynamical Model Evaluation System (DYMES) during the 2003 hurricane season in the Atlantic and eastern Pacific basins. Using the terminology of the Systematic Approach to TC Track Forecasting, I shall present the error mechanisms that resulted in large errors for the following five dynamical model tracks: NCEP Global Forecast System (GFS) and Geophysical Fluid Dynamics Lab (GFDL) models, U. S. Navy Operational Global Atmospheric Prediction System (NOGAPS) and Navy version of the GFDL model (called GFDN), and the UK Met Office global model. Initial findings indicate that track differences were often directly related to different depictions of the structure, especially intensity, of the TC. In particular, the TC in the GFS was often depicted as much weaker than in other models. How this difference affected the resultant track errors varied from case to case. Another initial finding was the high number of forecasts, especially for the UK Met Office model, that were degraded by a bad tracker. Other aspects of the season evaluation will be summarized.