The accuracy of dynamic stress-strain behavior for steel plate depends on how experimental technique is utilized properly. In the one bar method, one of the standardized high velocity tensile testing techniques, unfortunately, there is an apparent effect of the output bar supporting method on high strain rate tensile behavior of steel plate. If there is a little misalignment of the loading end of the output bar, lateral vibration of the end leads to an extraordinary-high initial stress peak on dynamic stress-strain curve. The author have been tried to find the optimum supporting condition of the metallic output bar end. Briefly speaking, at least one end of the plate specimen should be designed to be able to rotate in order to release inplane bending due to the misalignment of the bar, and no support or simple support with a V-shaped metallic support should be installed in the testing machine. However, these conditions were not enough to satisfy good reproducibility of the shape of dynamic stress-strain curves for the same specimen material. In this study, with a collar and a sheet made of polytetrafluoroethylene, the sliding energy at the contact area between the output bar and the supporting stand is tried to reduce. Good reproducibility has not been accomplished yet. Obtained experimental results suggest that some amount of contact pre-load should be introduced from the outer surface of the collar to reduce the lateral vibration of the output bar.