A format is quite easy to recover because it changes the file system structure, but not the actual contents of the disk.
Overwriting the disk with a homogenous signal is not quite safe, because you present a curious investigator with a known final state, so any differences found in e.g. a STM/AFM investigation (which is extremely slow and rather costly, btw) can be attributed to you "current" date.
This is to a lesser extent also true if you fill the HDD with e.g. cat /dev/random > $file until overflow. This does make the process significantly more difficult, though.
However, if you overwrite the disk with multiple passes of random data, there is no good way to atttribute differences found at the domain level to any state. Therefore, this is quite safe. If you do multiple passes with zeros, you simply make the differences smaller but do not make them go away.
Personally, I simply encrypt my more or less "interesting" data (which is mandated by my emplyer's rules anyway) and do one overwrite with zeros, which will thwart all non-professionals. And if the latter are interested in my letters to my grandma... well, then they can waste their time reading them
