Charged excitations of the oligoacene family of molecules, relevant for astrophysics and technological applications, are widely studied and therefore provide an excellent system for benchmarking theoretical methods. In this work, we evaluate the performance of many-body perturbation theory within the GW approximation relative to new high-quality CCSD(T) reference data for charged excitations of the acenes. We compare GW calculations with a number of hybrid density functional theory starting points and with eigenvalue self-consistency. Special focus is given to elucidating the trend of GW-predicted excitations with molecule length increasing from benzene to hexacene. We find that GW calculations with starting points based on an optimally tuned range-separated hybrid (OTRSH) density functional and eigenvalue self-consistency can yield quantitative ionization potentials for the acenes. However, for larger acenes, the predicted electron affinities can deviate considerably from reference values. Our work paves the way for predictive and cost-effective GW calculations of charged excitations of molecules and identifies certain limitations of current GW methods used in practice for larger molecules.