As self-assembled nanotechnology tackles increasingly complicated structures, biological self-assembly can teach us a great deal about the challenges of more complicated self-assemblies relative to the simpler systems accessible in current practice. The present study uses computer simulations of spherical assemblies inspired by virus capsids to understand the challenges artificial self-assembly systems will face as they approach biological levels of complexity. We quantify system complexity by two parameters-the total size of the completed structure in assembly monomers and the size of the first stable assembly nucleus. Simulations on a set of five model systems capturing a range of values for both parameters reveal several obstacles to extrapolating experience with simple systems to more complex ones. Assemblies of greater size result in total yields and assembly fidelities that are substantially more sensitive to the system parameters of intersubunit binding rates and to concentrations than are those of simpler assemblies. Larger nuclei partially mitigate these effects. Conversely, large assemblies have overall assembly rates with reduced sensitivity to system parameters, a feature that is also only partly mitigated by large nuclei. These changes can be partially understood by theoretical models based on nucleation processes, but such theory itself becomes less informative for the larger systems. We close with a consideration of mechanisms by which these obstacles may be overcome in actual viral systems.