One way to quantify a college football program's potential for success is to assess its probability of drawing in top recruits. Much of recruiting potential is dictated by the coaching staff and a school's prestige but geography plays a large role as well. The easiest way to observe this effect is to check out any team's roster. Both public and private schools will fill the bulk of their roster from students within a few hundred mile radius (e.g. 50% of players live within 250 miles of their school; see more details in "Methods" section).
Therefore a school's recruiting pool can be quantified by the population of the surrounding areas; in other words if we know how far away recruits are willing to travel, an effective population can be calculated for any location in the country (if you know the point sources of all populations). It's also important to look at the effect of schools competing for recruits, that is a school might have access to many recruits but these recruits might have offers from other nearby colleges. In this way schools were treated as negative populations.
In this way the 120 Division I-A schools can be ranked by their effective populations and places without schools can also be shown to be ideal places for a new program to crop up (theoretically). Schools can be also be shown to be over or underperformers.
More details on the mathematics used are located in the Methods section while different Cases/Analyses are listed below.
A few notes to start: for simplicity Hawai'i was essentialy ignored in this analysis although it appears in the results. The maps shown throughout use a rainbow scaling (the range is shown by a scale bar in the bottom right of each map) to show effective recruiting populations (in the saturation maps a red gradient is used to show high (redder) and low (whiter) populations). White circles are the sites of colleges. The numbers 1 - 10 indicate the highest population spots (all spots must be 30 miles away from each other. Some even end up in the ocean). There are some wacky looking lines on the map (e.g. the southern tip of Florida); that's simply due to the map-drawing code. I choose to use the more sensical "I-A" designation rather than "FBS" throughout the site.
Finally, you might find the pages to be a bit on the wide side (turns out the US is pretty big) as I wanted to preserve the high resolution of the figures, zooming out is often helpful to get the sense of the various map figures.
Much more will be forthcoming, check back for updates!
March 25, 2010: Added Case 3-6 which adds a multiplier for in-state recruits and a separate multiplier for private schools nationally.
March 24, 2010: Added Case 3-5 which adds a "Prestige" value from each team, check the case to see what quantity I used for the unquantifiable.
March 23, 2010: So much for the frequent updating--finally added Case 3-4 which adds a second winning percentage condition by also considering a college's WP over the three seasons prior to recruits' decisions.
January 31, 2010: Modified Case 3-3 to enhance (rather than dilute) the effect of winning percentage: Case 3-3.2. I have a backlog of data, so expect a few more cases to be posted soon.
December 21, 2009: Added two more normalized probability studies. The first, Case 3-2, is the analogue to the recently added study using winning percentage over the past 10 years as a scaling factor. The second, Case 3-3, also scales by winning percentage but also scales non-Big 6 teams to 0 (with the exception of Notre Dame) and yields much more realistic results.
December 16, 2009: Added Case 3, Sub-analysis 2 in which I scale the competitive effect by the teams winning percentage over the past ten years.
December 14, 2009: Added a page with winning percentages of teams each year since 1995. I'm currently working on using this as a scaling factor (analyses with that should be up in the coming days).
December 11, 2009: Added a Summary Page that should give you an as-brief-as-I-could overview of my methods and results. Many thanks to the link today from ESPN's Pac-10 blogger Ted Miller.
December 10, 2009: Added the first study using the normalized probability method using the Rivals 250 rosters and scaling by conference (the analogue to Case 3-1). On a related note, I added a handy table to the Population Sources page which shows the number of recruits from the Rivals 250 who went to each school in each of the past four years.
December 9, 2009: Added the simplest maps of all--the actual locations of the players/recruits/counties that I use for the various cases to the Population Sources page. Also added a page to the methodology section on using normalized probabilities to rank schools which I alluded to last week. The actual results should be posted over the next few days.
December 8, 2009: So I switched the style of the maps. Previously they represented populations by redness (or adjusting the "saturation" of the color). I altered them to vary by "hue" instead. I think these maps are much better at showing contrast (and they're nicer looking too), small downside is the file size is somewhat larger. I still have pages with the saturation maps as well so you can see the difference between the two methods.
December 3, 2009: Added Case 3 in which I use the Rivals 250 lists from 2006 - 2009 as the population source and the corresponding Case 3 Sub-analysis 1. I'm currently working on a different algorithm to help account for the effect of states with large land area on the map by adding each recruit to a specific schools based on his probability of going to a given school versus all other schools. This should remove the bias against schools like Arizona where the recruits may be far away but there aren't any schools closer. This will actually be a much simpler calculation to perform (but won't give a nice map to aid visualization or tell where the best places are to build a new program).
November 19, 2009: Added Case 2, Sub-analysis 1 in which I scale the competitive effect from other schools based on whether or not they are in one of the Big 6 conferences.
Calculating Effective Recruiting Populations
Calculating Normalized Probabilities For Recruits
Winning Percentages by Season for All Teams
In this case I use the 2000 Census to populate the map. The US population at this point was about 281 million.
In this case I consider each of the ~13000 players on rosters this year as populations of 1 to draw the map. (saturation version)
In this case I use the last four years of the Rivals 250 rankings to populate the map. (saturation version)