In reply to: Oh, those ubiquitous ratings! posted by BabaGhanouj
I think I understand point #2. Am I right in stating the proposition as perhaps, as players are rated farther from the top, the differentiation is more difficult and thus the ratings may be less accurate? Sounds reasonable. I agree it could be developed by more testing.
Point #3, I agree would be an interesting exercise. I would guess injury would be a big factor along with ones you mention. But that is untested.
If you consider the standard bell-shaped curve, there are fewer people at the extremes and they stand out more. As you move to the next levels, there are more people and it's much harder to differentiate. As examples. it's easier to name a first-team All America squad than the fourth team; or draftniks agree on the top 5 NFL draftees and disagree on the next 20 because there's not as much to pick between them.