That was part of the point I was trying make. The "round number" groupings do some weird stuff to the stats. The dataset I was looking at puts the 90% cutoff at 35 yards (a weird crater of 86% at 34 yards, makes me wonder if there is some weird likelyhood of kicking 34 yarders as game winners? Or just a funny stat).
In addition of those kicks being easier, because they are shorter, that's 18 possible kicking distances. Even "inside the 40" is skewed by them, because that's only 5 more distances vs. the first 18. "Inside the 45" is similarly skewed for the same reason: 18 possible distances vs. 10.
If we want to be human and love our base10 groupings, it might be better to do groups of 10s starting from 18 (1 yard line, 10 yard endzone, long snap: shortest possible FG is 18 yards) ala:
18-27 (100-97%)
28-37 (95-86%)
38-47 (83-75)
48-57 (68-54)
I don't like that, because in the second grouping, we already have a pretty wide variance. Grouping by 5s seems better, but that could just due to a finer grouping.
Counter proposal that is at best half-baked: We should instead rate kickers by what distance is their 95, 90, 85, and 80 percent accuracy mark is. The 10 year dataset would say 33 yards, 35 yards, 37 yards, 40 yards, and 42 yards. By drawing lines, it'd be easier to say if a kicker is better than average and add the qualifier of at what distances.
Of course we'd want a real statistician to draw those lines.