I just did some data mining on the stats from 2013. Nothing too exotic, just a bit of k-means clustering based on the 2013 defensive stats. http://www.pro-football-reference.com allows us to copy and paste the entire set of defensive stats at once for a whole year. So I just ran the stuff through my k means clustering program. My assumption is that the more clusters, the more the different defensive positions and defensive positions skill levels will distinguish themselves.
I realize though that simple clustering won’t necessarily know a LB from a DE or a S, but good stats are good stats so I just wanted to see who the program paired together. I’ll see if I can later find a better set of stats to play with.
Check it out