Population Sources

There are multiple ways to define a recruiting base. It could be based on pure population. It could be based on the number of athletes from a given area already playing, or it could be "good" recruits, e.g. those featured in the Rivals 250 or a similar recruiting rating service. You could even scale recruits by the number of stars they've been given by the various services.

Each of these approaches is used on the site and the differences between them are interesting to note. When comparing the entire population to the population playing college football the largest anomaly (unsurprisingly) is in the northeast--the large population from New York to New England does not translate to many football players. The entire population data sources come from this 2000 census site where the US government has kindly posted the data files contained therein. For ease of computation, the counties file rather than the cities file (~25,000 versus ~3,000 populations to load) was used (see Figure 1 for the locations of these counties).

A second population source are the players on the current team rosters. Each player's hometown (as published in their rosters on espn.com) was placed on the map with a population of 1 (see Figure 2 for their locations). For the more numerical details on how these populations were handled, see the Calculating Population section

One very interesting question is whether the ~13000 players on I-A teams are in fact the best 13000 players in the country. In other words, do teams' geographies actually create recruits simply due to the ease of recruiting locally? I imagine in the pre-internet days this was actually a valid concern, i.e. many players were going unnoticed simply because they live too far from major programs. Given the national scope of recruiting and the availability of scores and video clips via the internet this effect should be minimized. The other ignored data are players at lower levels (I-AA, D-II, D-III) who have the talent to play for D-I-A schools. Given the large sample size of I-A players, I believe this does not significantly affect the data.

For Case 3 the combined Rivals 250 lists from 2006 - 2009 (i.e. those players liable to be on rosters in the current season, 2009) (the lists can be found here) were used as the population source (Figure 3). 929 of the 999 (2006 had only 249 for some reason) players were found via the coordinate lookup script. As with using all players on rosters, each prized recruit was given a population of 1. Below Figure 3 is a table indicating the number of players from the Rivals 250 who went to a given college each year. Unlisted colleges did not attract a Rivals 250 player from 2006-2009.

Rivals 250 College Choices, 2006-2009.

Note: "Scaled column" has the total multiplied by 929/999 (the percentage of recruits located on the map)