We previously developed Upside, a near-atomic, fast molecular dynamics algorithm for protein folding. A key feature of the model’s efficiency is the representation of sidechains as single coarse grained beads and a rapid calculation of their rotamer free energies for each time-step, giving a smoother energy surface for the backbone to evolve on. We used the contrastive divergence technique from machine learning to train from simulations of 450 proteins for which our model’s efficiency allows for better representations of the Boltzmann ensembles for precise tuning and greater accuracy. The model is afterward able to de novo fold proteins up to 100 residues on a single core in days.Here we were inspired by Upside’s folding performance to adapt the model to predict protein-protein binding. Predicting protein binding is a core problem of computational biophysics. That this objective can be partly achieved with some amount of success using docking algorithms based on rigid protein models is remarkable, although going further requires considering the effect of protein flexibility. However, accurately capturing the conformational changes of the proteins upon binding remains an enduring challenge for docking algorithms. We use Upside to investigate when backbone flexibility helps docking predictions, what types of interactions are important, and what is the impact of coarse-graining on accuracy. These efforts also shed new light on the relative challenges posed by folding and docking. After training the Upside potential for docking, the model is competitive with established methods, but with some loss of accuracy due to the absence of atomistic side chains. Allowing for backbone flexibility during docking appears to be generally detrimental, as the presence of comparatively minor (3-5 Å) deviations relative to the native folded structure has a negative effect on performance. While this issue appears to be inherent to current forcefield-guided flexible docking methods, antibody-antigen complexes represent a major exception. These systems involve the co-folding of flexible loops that benefit from Upside’s backbone flexibility.