Were I to graph the calories burned per week exercising for group in my social network (were I to bother maintaining my Facebook page), it would create a normal distribution. This graph would probably correlate pretty directly to their perceived fitness and their enthusiasm for exercising. Within this group I could find a subgroup that was slightly-more-enthusiastic-than-average (and subgroups that were slightly-less- and considerably-more- etc.).
If I had a New Year's resolution to burn more calories per week exercising, I could compare myself with a subgroup whose mean calories/week was slightly more than my average. This would be more effective than comparing myself to my most-fit friends, whose enthusiasm and dedication to fitness is as unreachable as the skills of teenagers in Halo 3 (note clever allusion to TrueSkill).
As time went on and my New Year's Resolution waxed (or waned), the group used to determine my subgroup would shift (hopefully toward the more fit, but always to "slightly more motivated than your current evidence"). Note that it's possible for essentially everyone to be in such a group, given a large enough set, i.e., the set of people who are interested in getting more fit.
I understand there's money to be made in Facebook applications. Does such an application exist?