I kicked this off without benefit of any real world context but there ARE some things to be considered.

Such as, we are stuck with limited data so are trying to find some proxy or estimate for actual measures of "useful torque," that would require the kinds of data Tak describes. We know that "anything else is just guessing" but don't have any choice but to do so.

Ultimately, the questions are DO WE guess and, if so, HOW...? **

K

** By "guess" I mean "informed guess," not POOMA. We're trying to get away from the latter. EDIT - and we're talking about informed guesses to establish repeatable practices/processes for all cars, rather than informed guesses in isolation for each car we look at.