Despite the widespread use and numerous successful applications of density functional theory, descriptions of hydrocarbon reaction energies remain problematic. Illustrative examples include large underestimation of energies associated with alkane bond separation reactions and poor general description of intramolecular dispersion in hydrocarbons (e.g., B3LYP, MAD = 14.1 kcal mol-1). More recent, but not readily availably functionals, along with efficient posteriori corrections, not only show considerable improvement in the energy description of hydrocarbons but also help identify the sources of error in traditional DFT. Interactions in branched alkanes and compact hydrocarbons are adequately mimicked by systems compressed below their typical van der Waals distances. At these distances, standard DFT exchange functionals are overly repulsive for non-bonded density overlaps, and significant improvement is offered by the long-range corrected exchange functionals (e.g., LC-BLYP0.33, MAD = 5.5 kcal mol-1). For those systems, the neglect of long-range dispersion is found to be a critical shortcoming, as well as ''overlap dispersion'', for which non-negligible amounts are captured by the correlation functional. Accounting for the missing dispersion interactions is of key importance. Accordingly, most noteworthy improvements over standard functionals are obtained by using non-local van der Waals density functionals (e.g., LC-S-VV09, MAD = 3.6 kcal mol-1 , rPW86-VV09, MAD = 5.8 kcal mol-1), a dispersion corrected double hybrid (B2PLYP-D, MAD = 2.5 kcal mol-1), or by the addition of an atom pairwise densitydependent dispersion correction to a standard functional (e.g., PBE-dDXDM, MAD = 0.8 kcal mol-1). To a lesser extent, the reduction of the delocalization error (e.g., MCY3, MAD = 6.3 kcal mol-1) or careful parameter fitting (e.g., M06-2X, MAD = 5.6 kcal mol-1) also lowers the errors.