Remember that there’s almost certainly “skew” in the “top 100” scores (statistically speaking) - which just means the median will be lower than the mean. Which means that if the cutoff was exactly the mean of 216.6, you would get a lot less than 50 kids meeting that cutoff. Maybe 30? Who knows - just a guess. It’s possible to construct a (not realistic) distribution where only 10 would make it - 40x214, 30x215, 20x216, and 10x228 would average 216.1, but only 10 would have higher than the 216.1 mean.
No worries; I understood the intent behind it. Sometimes it’s hard to convey tone through writing. 
I thought roided up was a funny term for it, actually, because it’s super ironic.
Knowing the area, we do see flocks of uber smart kids suddenly appearing at the latest hot public school, and numbers can go haywire within one year. At my kids’ school, their grades jumped by about 200 kids each (junior and soph), and the freshman class had a whopping 280 more kids this year than the year before.
That’s how Atlanta rolls, unfortunately. Everyone hears about the “good” school and moves to that district because they’re out of state, been transferred by their company, and listen to their real estate agent about avoiding the bad schools, and then they kill the golden goose through overcrowding.
It’s one of the downsides of Atlanta being highly mobile and tech-y.
It’s interesting to see a statistical analysis of how students score. It’d be curious to see if this same kind of mechanism can be used to generate some kind of automated question selection algorithm for adaptive testing/testprep in the future. Even a port-mortem for detecting cheating, structural discrimination, or faulty testing sites would be interesting.