Accolades and feats sort of are the same thing when you peel back the layers.
A feat comes due to an event described by a narrator, .ie, it's an accolade.
An accolade itself is a feat - the feat of impressing a character or narrator.
In either case, there are a few criteria through which we can compare the reliability of feats vs. accolades:
1. Consistency/predictive ability. How neatly can you sort feats together to form a coherent narrative, compared to accolades? Here, accolades clearly win out.
2. Suspension of disbelief/reality. Which model best allows us to construct a universe that feels real and therefore to make such questions seem relevant? Here, we'd go with feats.
3. Authorial intent. Which model would authors pay more attention to or give more thought on? Accolades win out here.
4. Quantification. Which model lets us deduce these Force users' abilities from an absolute framework, rather than just relative to one another? This is particularly important to crossover vs. debates. Here feats have the clear advantage.
5. Enjoyment. Which method of debating is more fun? This sounds like an arbitrary standard, but since there's no objective reality with respect to this imagined universe anyway, it's the ultimate end of these debates. I guess this is a rather subjective question.
Although obviously you want some combination, probably based on the extent to which it affects the probability that X character is more potent than Y, and/or the extent to which it makes the model that X character > Y the most parsimonious one.