But you can't draw any conclusions from your sample without understanding how the health ratings are distributed in the sample.  You basically seem to have assumed that since the ratings go from 0-100, 20% of all players have a health rating above 80 - and that's just not true.  The health rating distribution is pretty heavily waited toward the higher end of the scale.

I looked at the most recent draft classes in my two leagues.  (I have 20/20 scouting in both) In both cases, fully half of the top 500 prospects project to 80 or better health, and 1/3 of the players were already over that mark.   Less than 30 have projections of less than 50 for health.  Now, obviously players fall short of projections, but if 1/2 of players being generated project to 80+ health and 1/3 start over that mark, I think it's a pretty fair assumption that significantly more than 20% of players have a rating above 80.  If, say, 40% of all players are over 80, wouldn't your results demonstrate that health DOES matter?