Checked with the sage-like D&D rules forum, I'm adding +1 CR to true dragons for sure now. Having done that, the average HP notches down to 54, the standard deviation goes to 15. Outliers lower than 39 or higher than 69 become:
Very Young Blue Dragon, Orca (high)
Adult Arrowhawk (by 1 point), Rast, Ravid, Shadow Mastiff, Wraith, and Unbodied (low).
Interesting... I read the articles, they're definitely related to what I'm trying to do here (but possibly the more reasonable focus on fixing the car rather than making the bicycle look like one).
Did you measure damage output? At low levels, some creatures have disproportionate effect due to Str. Granted, that might be a corner case in your analysis. Anyway, cool stuff and thanks for sharing!
I have done some work before on damage, but I haven't done the stats work yet because parsing the "attack" line is tough. When I do it, it will be average damage output on an attack. A GM who wants to do a "damage roll" off that can use the hot die, cold die method to add some variance.
Here's the result table. I just put Average and Standard Deviation (add it to get "high", subtract it to get "low"). This isn't the pretty table, this is just the "hey, here's the concepts" table.
Yep, that's intentional. The high and low are supposed to create a "normal" range (not normal in the statistics sense, just ones that are relatively typical). See the stuff I posted about outliers, above.
Here's the second batch, also includes Max and Average Stats, and Max and Average Saves.
Really ugly though. But this is nearly all the info a GM would need (DC of spells notwithstanding). Of course, I'll smooth these into curves before I make my final chart.