I have to admit, I didn't expect the log regression coefficient between the weighted grade averages (Grades from HF) and round# to do better than the one I did for the baseball draft last year (which had MLB players, not prospects). What's more surprising is that this draft was done before most of the updates were made on the site.
Methodology was simple. I weighted the HF grades (1-10) by their risk (A-D,F), the latter of which was weighted by HF grade (lower grade = greater risk). Also, I adjusted for position (G = 2/3 F weight; D = 1/2 F weight). It's surprising that something done with what I found to be arbitrary means of weighting did so well (I guess, since it's relative and not absolute, it makes such arbitrary weights easier to do), as well as the fact that this is coming from Hockey's Future's grading system!
Here's the plot based on average weighted grade per round:
http://oi59.tinypic.com/xm3n6o.jpg
Here's the plot based on individual weighted grade per pick:
http://oi58.tinypic.com/24ya4pg.jpg