How rare the event, and the consequence of the event are the two equally important factors. You can't decide on the basis of either, without the other.

A case in point is the article showing up recently in this thread, where cyclists with a traumatic brain injury were about 50% more likely to have a more serious one (or a fatality) without a helmet. That is the correct interpretation - please don't go off on a tangent on how statistics cannot predict the probability of an event for an individual. I'll register your objection in advance. However, as extreme as the consequence may be, the article neglected to tell us what the chances of the TBI is in the first place. If there was a one in a billion chance of suffering the TBI on a bike ride, then that 50% risk of greater injury is mostly irrelevant. A rational person wouldn't bother with a helmet in other words. If it was one in a thousand chance on the other hand, any rational person would see that the avoidable 50% chance of greater injury is too much risk.

Since I've begun riding, I have a fall of some sort every 5,000 to 12,000 miles, trending to the lower frequency over time. Since the basic falling and rolling skills not only prevent head injury but also can reduce or eliminate broken bones and minor injuries in each of these falls, it is unquestionably worthwhile in my estimation.

These skills are more common than people perhaps realize, and perhaps that's why my question hasn't been formally asked in these studies. Nor answered here.