Trying to use estimated miles to measure pack capacity would be useless since estimated miles depend on recent driving which can't easily be controlled across time or owners.
Rated miles and ideal miles are both energy units, and thus a good measure of battery pack capacity after a charge. To get the best reading for a Roadster, you have to wait at least 10 minutes after the charge completes for the Roadster to make additional measurements and refine the SOC estimate. I suspect it's similar on the Model S.
The rated/ideal miles vary too much to be of much use, as the Model S is too new to have much more battery degredation than the small differences reported. Then there are the differences in firmware that change the reported numbers. A good example of this would be my car. First full standard charge was 237. Then it varied berween 239-241. Then after a while it went back to 238. A few charges back I saw 245. Now its 240. This is on the same firmware 4.2 that the car came with. This data tells me that there are too many variables, and trying to come to a conclusion about battery degredation from this would be silly at best.
I have noticed that some here go the opposite direction I did(start out high and go lower), which in turn makes them freak out and cry wolf. A few miles down after a few months isn't going to be a conclusive way to determine battery capacity.
On the Roadster, we can get the "calculated amp-hour capacity" of the battery pack, which Tesla says is the best measure of the pack's capacity. This number is much easier to collect than reading the ideal range after a charge given the care that must be used to get an accurate reading, especially for a range mode charge where you have to wait at least 10 minutes for the SOC to settle, but don't want to wait any longer as the car slowly uses energy from the pack to continue thermal management long after the charge completes. That said, I've found that standard mode ideal range correlates very well with CAC. Range mode also correlates, although with a lower correlation coefficient, which I attribute to the logistical challenges in getting a perfect range mode reading.
Of course all of these numbers are ultimately computed by the car and susceptible to change and even deliberate manipulation by the automaker. The only truly reliable way to measure pack capacity is to run the car from full to empty in perfect, standardized conditions, which means careful charging procedures followed by running the car on a dynomometer under perfectly controlled conditions. That's an expensive undertaking and lot to ask for a widespread, grassroots survey. We can do almost as well with simpler data, combined with owners logging trips to compare rated mile performance under similar driving conditions across time.
In a very real sense, perception is what matters. Helping Model S owners relate what they see to what other owners are experiencing would be helpful to the community even if the numbers aren't as scientific as we'd like. This played out in a big way for the LEAF community where the car's instrumentation is incredibly weak.
I think it would make sense to start a study based on rated miles observed after standard and range mode charges. The way I have the survey data collection set up for the Roadster and LEAF surveys makes it's easy to see both a snapshot of
the most recent reports from all vehicles and
the history of reports for a given vehicle.
If a better way of measuring battery pack capacity is made available later, that can be added to the survey and correlated with the history of reported rated range values.
I'd be happy to provide a web API way to submit reports to enable apps that could allow owners to use the Model S REST API to securely grab data for their vehicle, then submit the relevant info to the survey. That way owners don't have to share their passwords with anyone but can still submit data in an automated way.
This might be the best way to go about it