We have developed a tracking algorithm to determine the speeds of supra-arcade downflows (SADs) and set up a system to automatically track SADs and measure some interesting parameters. By conducting an analysis of six flares observed by the Atmospheric Imaging Assembly on the Solar Dynamics Observatory, we detect more smaller and slower SADs than prior work, due to the higher spatial resolution of our observational data. The inclusion of these events with smaller and slower SADs directly results in lower median velocities and widths than in prior work, but the fitted distributions and evolutions of the parameters still show good consistency with prior work. The observed distributions of the widths, speeds, and lifetimes of SADs are consistent with log-normal distributions, indicating that random and unstable processes are responsible for generating SADs during solar eruptions. Also, we find that the fastest SADs occur at approximately the middle of the height ranges. The number of SADs in each image versus time shows that there are rest phases of SADs, when few SADs are seen. These findings support the idea that SADs originate from a fluid instability. We compare our results with a numerical simulation that generates SADs using a mixture of the Rayleigh-Taylor instability and the Richtmyer-Meshkov instability, and find that the simulation generates quantities that are consistent with our observational results.