So I've done enough 'journeys' of my most common daily drives (commute plus after work trailhead and then back home) to have a decent sample for an analysis of how much energy is spent going uphill vs how much gets regenerated on the downhill. Here's what I did.
I calculated how much energy I expect to spend extra on the uphill (or get back on the downhill) for all the segments I drive. Energy required for lifting is a simple calculation, E = mgh where m is mass, g is the earth's gravity, and h is the height in meters. Comes out as Joules, divide by 3,600,000 and voila you have the kWh for the h, the elevation delta of the drive.
Then I subtracted that figure from the energy used for the whole drive. In other words, I "factored out" the predicted contribution from the climb or the descent. (For the descent I added the predicted amount of energy gained, which is the same as uphill times -1.)
The remaining energy consumed is that used for propelling the car forward against the air and other consumers, as well as drivetrain inefficiencies, capriciousness of driving, etc.
Here's what I found. See the attached thumbnail. (I forced the intercept through zero but it made virtually no difference.) Distance driven explains 95% of the variance in energy used. In other words: uphill slope, downhill slope, a bit more spirited driving, a few more traffic lights, many other 'confounders' you can think of ... contribute only 5%.
The conditions for this test were pretty favorable to get a strong correlation like this. Always comfort mode; hardly used climate (heated seats here and there), it's the same route several times over, temperatures did not vary wildly. The one outlier (the rightmost point at about 11.2 miles) was a cold day, hence a bit more energy used. No difference between daytime and nighttime.
I think this is pretty wild. I suspect the same is true for Teslas but I wanted to see it for myself and haven't scoured the web for a similar analysis yet. At any rate, this vehicle is an engineering marvel and possibly the only thing you can hold against it is the slightly suboptimal drag coefficient. I'll take the resulting looks any day for that price.