I agree they can't be compared with any other numbers, but can't they be compared with themselves? If the same methodology is used and the rates are trending down, that implies progress no?
Assuming they haven’t changed the categories somehow (and we would have no way of knowing), yes, you could track the safety of each category relative to itself over time. Obviously only once a year, since there are seasonal variations. You also have to account for the demographics (and location!!!) of Tesla owners changing over time (again, only Tesla has this data), and how that impacts the safety. For example, a lot more owners in winter climates (or more hazardous locations for whatever reason) should perhaps lead to a higher accident rate with time.
So, it’s really hard to make anything other than vague determinations without the entire dataset! Unless of course accident rates go to zero, or get very very low (that would mean a lot).
I guess they’ll have to add a HW3/FSD category!