To assess the accuracy of post-Hartree-Fock methods like CCSD(T), MP3, MP2.5, MP2, SCS-MP2, SOS-MP2, and DFT-SAPT, we evaluated several effects going beyond valence-correlated CCSD(T). For 16 small hydrogen bonded systems, CCSD(T) achieves an RMS error of 0.17 kJ/mol in the dissociation energy compared to our best estimate, which is a composite method akin to W4 theory. The error of CCSD(T) is thus much lower than for atomization energies. MP2 is surprisingly accurate for these systems with an RMS error of 1.3 kJ/mol. MP2.5 yields a clear improvement over MP2 (RMS of 0.5 kJ/mol) but still has an error about 3 times as large as CCSD(T) for the absolute RMS and almost 10 times as large for the relative RMS error. Neither SCS-MP2, SOS-MP2, nor DFT-SAPT yield lower errors than MP2. With a ΔCCSD(T) correction to MP2, the basis set limit is readily achieved when employing diffuse functions-without these, the convergence is rather slow.