Structure-based computational design techniques can significantly reduce the time and expenditure required for new drug discovery. However, origins of ligand binding which drive molecular recognition are difficult to predict and remain a challenging problem. Docking and virtual screening have proven to be very useful computational methods but further development, testing, and algorithmic improvements are needed to make such techniques more robust, reproducible, and ultimately more fruitful. In this study, we have developed a testset consisting of ~600 hand-curated x-ray small molecule ligand-receptor complexes, constructed from the Protein Databank, for use in redocking experiments. The goal is to evaluate different protocols for recovery of the bound ligand pose, using the program DOCK, and examine causes for failure. Results will be presented for redocking the entire set as well as an examination of docking successes for individual protein families including carbonic anhydrase, alpha and beta trypsin, neuraminidase, thrombin, T4 lysozyme and coagulation factor Xa. The correlation between binding energies estimated with DOCK and available experimental activities will also be reported.