Loudness Penalty Tricks

[pftq]

More for my own notes for reducing loudness penalty of music but some of these took forever to figure out - maybe helps some others out as well.
- Loudness penalty is based on the average loudness of the track.  So yes, having quiet parts reduces your average and boosts the overall loudness.  However, things that are too quiet (like actual silence) are automatically ignored and not counted toward the average.
- Stereo widening instantly worsens your loudness penalty even if it's perceptively the same volume.  Conversely stereo merging instantly improves your loudness by 1-2db.  This one took me so long to realize - I was EQing a track to death only to realize it was because of how far left and right I had things panned.  Similarly drums tend to be stereo merged in the lower frequencies for commercially produced tracks, and this is probably why.  Vocals and lead synths/instruments tend to be entirely mono to begin with, which probably is this same thing.
-  Counterintuitively (because bass is instead usually the problem in mastering/mixing), the area around 2000hz is actually what's penalized most, since that's the most audible for human ears.  Additionally, for someone like me who loves having too much stuff in the highs, strings/choirs can be dropped a db or two without perceptively sounding that different but it does a lot to reduce loudness penalty.
- Lastly loudness penalty is really a composition/mixing problem, not mastering.  Don't try to regain penalty points by EQ-ing out frequencies; it'll just end up sounding thinned out or over-mastered.  It matters more to vary the volume/velocity of instruments more and add actual quieter parts of the song.