"Automatic based on current playlist" uses the ReplayGain data to bring all tracks to a normalized level, and then plays the current playlist as loud as possible without clipping.
This means you will have to adjust the volume control when you start playback (because it varies depending on what is playing) but after that, you should not have to touch it for the rest of the current playlist.
If you are using album-based ReplayGain without any adjustment, then all tracks will be normalized to 83dB. I'm not sure on the specifics as to why 83dB was chosen for Media Center, as the
ReplayGain specification is 89dB.
If you want that, you will have to use a +6dB fixed adjustment. I would not set it any higher than that, as you then run the risk of clipping or activating Media Center's clipping protection.
While it might sound quiet compared to playing tracks without ReplayGain, the whole point of operating in this mode is that 83dB gives you 20dB of headroom (I don't understand how -20dBFS = 83dB though?) so that you aren't likely to run into clipping when normalizing the audio levels.
Once you adjust the volume control on your amplifier to a comfortable listening level, you should ideally never have to touch it again, because all tracks are being normalized to the same level rather than it varying based on what is in the current playlist.
In practice, it's not quite perfect, and I do still have to adjust my volume control up and down a little, but far less than I would without using ReplayGain.
Normalizing this way also defeats the "loudness war" because all tracks are now being played at the same level, so what you hear is the difference in dynamic range, rather than the difference in loudness.
And looking to the future, I hope that Media Center will implement EBU R128 normalization which supersedes ReplayGain, and performs a more complex analysis of the audio to improve normalization results.
R128 normalizes at 80dB, leaving 23dB of headroom, so it will actually be equivalent to ReplayGain with a fixed adjustment of -3.0 dB.
If you are outputting a 24-bit signal from Media Center, you shouldn't really be concerned about this affecting playback quality at all, as 24-bit allows you to reduce the volume by up to 48dB with a 16-bit source. (which most music is)
And even if you have a 24-bit source, you were never going to hear anything down below -121dB anyway.
Track-based ReplayGain is useful for when you are using the Play Doctor, or more random/shuffled playlists, as it will normalize each track to the same level, rather than normalizing whole albums to a single level and preserving the track-to-track dynamics. Some tracks are supposed to be louder or quieter on the album than others for example, and track-based normalization eliminates this.
