I think it's a good idea for Obama and Clinton proponents to propose different measures for the purposes of argument to the superdelegates, like Bayh's test. But that test needs to be refined to focus on states that might go Republican in the fall. New York, California, and Massachusetts don't matter. You may think the fairest test really is the pledged delegates, because that is what the candidates were competing for. Bayh's test retroactively treats state primaries and caucuses as though they were winner-take-all, but if they were the candidates would have fought differently. But it's not clear to me that the candidates relied on something that the superdelegates have to respect.
The superdelegates should be using their power to choose the best candidate for the party, and there are many arguments that could refute the assumption that the candidate that collected the most delegates — but failed to lock the nomination — will do best in November. Since November is crucial, the Electoral College should be taken into account, but not in the crude way Bayh suggests, since the superdelegates shouldn't worry about marginal preferences for Obama or Clinton in states that will almost certainly give its electoral votes to whichever candidate the party nominates. Thus, Clinton's popularity in Ohio and Pennsylvania (and, face it, Florida) matters, but her popularity in New York and California doesn't.