Yes I did mean skill ratios, but it's not my point that the data is not showing what we've measured correctly, but wether we are measuring the correct things. We have data that shows us how our population is shaped, but when I decide what I want to train when, I don't need to know what the effect is on the most commonly trained player is, but what the effect on my player is. I think it's different if I don't have the most commonly trained player.
I'm not trying to argue what the exact difference between the two is, just that there is or could be a difference and that it matters. If the results also go against logic, that's more reason for me to believe there's a significant difference.