There are no doubt many theories on the macroscopic effects that the different steel phases have on etching, what I can give you are the actual effects that I have observed on the microscopic level, which correspond uncannily close to those same macroscopic effects.
Annealed steel doesn’t etch so well because if the annealing is thorough you have separated a maximum amount of carbide from the iron, leaving large fields of ferrite with large chunks of carbide. Ferrite, being as close to a perfectly uniform iron crystalline array as you can hope for in steel, etches very evenly and thus does not scatter light well, it just bounces it right back at you. Ferrite is always white under the microscope for this reason, and to the naked eye as well. If the anneal is lamellar (pearlite) then the light is scattered a bit more and you can get a mediocre etch, but if the anneal is spheroidal, whole grains of ferrite littered with shiny white cementite spheres just bounces light right back at the naked eye and the etch is the worst.
Fully hardened steel is a little less uniform since martensite is acicular (needle like) but untempered its needles look like mirror polished Christmas tinsel or garland, so the etch is better but still boring. Unlike spheroidal carbide really fine carbides are dark and widely scattered throughout the steel, and perhaps the finest and the darkest are tempering carbides. As you begin temper the steel the ever increasing formation of sub-microscopic tempering carbides begin to have an accumulative effect of darkening the martensite needles, tempered steel can be easily identified under the microscope because the martensite is darker in color. The more tempering done, the greater this effect, until one begins to form spheroidal carbide again.
What it all comes down to is how much light is reflected straight back at you making things shiny and how much is scattered in all directions, making it dark. The different phases will have greater or lesser scattering effect, while alloying will resist the etching more or less in producing the scattering. So a steel that is high in nickel, while still comprised of tempered martensite will resist the etchants ability to produce light scattering planes or facets. Manganese, on the other hand likes to gather between martensite laths or needles and accentuates the scattering effect.
