ygelman wrote:
We agree; that's why I called it a hum-drum mantra, and I like your style . . . but I'd still like to see an expert comparison.
Agreed. I think this is something each of us should do for our own reference purposes. Take a scene with a lot of range, add a 21-step gray scale and a ColorChecker chart, and perhaps a several races of live human models, and record both raw and large, finest JPEGs at 1/3 stop intervals over a +/- five stop range. Then mess with the raw images in post to see what works FOR YOUR PERSONAL STYLES. Evaluate the JPEGs, too, if you like, to compare the available range.
I haven't done this in a few years, and no longer have the files (they belonged to my former employer), but I did such tests a decade ago, when testing new cameras. It's a great exercise.
In general terms, what I learned was that ETTL works well if you use JPEGs... A 1/3 stop underexposure, as measured with an incident meter, is about right. ETTR works well if you use raw files. About one stop, ±1/3 stop, of overexposure greatly enhances shadows and still includes recoverable highlights in most average daylight scenes.
Raw files have so much data (which translates to latitude) that you can generally go a stop under to a stop over with few worries and subtle differences. Get very far outside that range, and you're either producing a special mood or effect, or you are getting sub-optimal results, depending upon your interpretation!
JPEGs generally require a dead-on accurate exposure at the camera to look their best. I've always said the exposure latitude at the camera is that of slide film about +1/3, -1/2 stop. Outside that range, you blow the highlights or lose noticeable shadow details.
Again, do what works for your interpretation of a subject.