Eight CAPRI prediction rounds with a total of 15 targets were held in the years 2010–2012. Only five of the targets were protein assemblies comparable with those of earlier CAPRI rounds. In one target, the solvent positions at the interface had to be predicted; another was a protein–polysaccharide complex. The remainders were designed complexes issued from protein engineering experiments, and the prediction concerned either their structure or the binding affinity of the designed ligand. Affinity prediction was a new experiment in CAPRI, and a challenge for its participants. It pushed the community into developing novel procedures and score functions that will improve the performance of docking methods, help designing binders, and yield better structure-based estimates of the binding free energy of natural assemblies. Proteins 2013; 81:2075–2081. © 2013 Wiley Periodicals, Inc.