This is actually something I've been meaning to bring up here for a while now...
I refurbished a
cabinet with a 20" vertical CRT a while back and fitted it out with some
refurbished LS-30 rotary sticks.
The games list was strictly vertical games with rotary joysticks (including Calibre .50 that uses the Loop 24 stick) so the games list was short.
Aside from the
hassle of getting the rotary sticks working, one of the things that really annoyed me once I had it up and running the volume discrepancies between the games.
Long story short, I wrote and AutoHotkey script that imported sounds recordings of each games (I played each one for 300 seconds) into
Foobar and then analysed the recordings to create
ReplayGain values. From there, my AHK scripts created individual *.ini files for each games with the ReplayGain value used to attenuate the volume level (attenuate because MAME doesn't provide a way to boost so the quietest games dictates the highest gain).
The end result is that all games have extremely similar average volumes levels now. I suppose increasing the length of the sample recording would make for more accurate analysis but I found 5 mins produced a good result. Of course, the downside is that I had to play every game to produce a WAV file to analyse (not such a terrible task) but it would be very time consuming if the games list wasn't so short. If we got organised and distributed the task though...

I could dig out my script (the final version is stored on the cab which is in a cafe set to freeplay) and post some instructions if there's any interest. It's a lot more accurate (and less frustrating) than trying to match the volumes by ear.