I don't get why one leaves gain headroom for mastering.
I hear it all the time so I know it must be important but..
If I have exported a stereo wave file to master, surely I can simply reduce the gain in one of several ways before adding EQ, Comp. or limiting etc.
What am I missing?

Reducing the gain will not restore the dynamic range.

Ah! so the point is not to squash it too much, not that you should keep the master level down a few db before export.

Just ask any professional Mastering engineer, and he'll tell you that, as well, but you may have read or heard this from a source that may not have stated the destination medium, such as vinyl or CD. 

For vinyl, you need to leave about 3db minimum of headroom. 
For CD, in my experience, you mainly just don't want any overs (no RED LEDs), but to mix at a good loud level, meanwhile retaining all of your dynamics. In other words, use no limiters or compressors on your Main buss. 

I've discussed this at great length with some major label mix engineers, and the common thread is that you should get your mix to "sound" like you want, and set the Master volume such that your maximum peak is less than 0db. From there, the Mastering Engineer will do the rest. 

My own recommendation in another thread is to talk to the Mastering Engineer you'll work with. Ask them. They will tell you exactly what they want most and what they want least.
 

Ive kinda wondered this myself. In the past I have recorded bands and applied nothing on the master fader. At the end, after mixdown, I would normalize it to make it as loud as possible for the client (once again with no extra processing on the master fader). But when I send these files to a mastering engineer they come back and say that they have no room to work and that I need to resend the songs. So I send then the same exact file without the normilazation and everything is ok. If a mastering engineer is just going to chop off the transients with a compressor to bring the overall RMS level up then why can't they work with the pre-normalized file as only the very single top peak of the entire normalized file is hitting the ceiling when I send it to them(and the song will be crushed to the ceiling by the ME anyway. Or why as stated by the OP why can't they just back it down a few DB if they need the room to start crushing? I hope this makes sense.

DON'T normalize before mastering. Give them 3-6db of room to work with. DON'T limit or compress on the master buss. No need to discuss  further. End of discussion. 


Lance

As a rule I leave -3 to -6 db headroom. However you should consult with the mastering engineer and find out what they want. I had a guy who wanted songs at -.1 (no kidding) so he could master them.

It's based on either completely incompetent mastering engineers, fear of communications problems or misunderstanding about desired output levels between mastering engineer and client, or not using* or understanding floating point bit formats.

Or else it's just people making a bad, confusing and misleading argument to try and achieve a good purpose - not overcompressing the crap out of everything.


*I don't understand why anyone in the modern world would use anything less than a 32bit fp delivered to mastering.

As I understand now, it's more a function of leaving dynamics (minimal compression and limiting)  than levels.
One could over comp and limit on the tracks and do the same damage.
Right?
So it seems we need some guide that indicates acceptable dynamics rather than levels at master.

"*I don't understand why anyone in the modern world would use anything less than a 32bit fp delivered to mastering. " to be fair most people probably do Drew