The power of DFT

 

"Fortuitous error cancellation" is the premise of DFT. Albeit, 

not only do two wrongs make a right; three can, and so forth.

In general, some "lower" levels of theory perform better for this; in Sherill's words,

these may represent so-called "Pauling Points" due to the aforementioned coincidence.

Alas, even given the exact same level of theory, results may vary across computational

software (e.g., Gaussian versus ORCA) for reasons that are nigh impossible to discern.

Sometimes, performing the exact same calculation in the exact same program will lead

to different results; JCIM mandated triplicate MD simulations for this reason, albeit they

are indeed a journal focused on methods development (where sub 1 kcal/mol "chemical

accuracy" doesn't quite cut it for many). So when espousing the wonders of DFT, as someone 

like Kieron Burke does, one ought to be careful, lest feed the "computational nihilism" 

that besets our field.