I'm attempting to teach myself HF and DFT by implementing various algorithms, but I've run into a snag: many molecular properties which are experimentally measurable appear to require taking gradients, which is considerable extra work (from a coding perspective). Are there any properties where I can start with a known optimized geometry, run my HF or DFT implementation, and then compare the results to experimental values? Pedagogically, I'd like to be able to motivate DFT by showing it clearly outperforms HF SCF in some experimental regime.
Many thermochemistry benchmarks, like the GMTKN55 (homepage listing most results, DOI), use fixed geometries and require only ground state total energies. Some of the subsets probably have back-corrected experimental references, but you would have to check each paper, I'm afraid. Note that for some properties, a dispersion correction is vital.