Recruitment database: 1980-2022 | The key game
If you want to play around with recruiting data going back to 1980, spreadsheets are currently available for download here:
This guy is an NC State fan with a hell of a passion for barbecuing and an amazing collection of rides on his page, too. Like, seriously, this guy does a lot of hiking and it’s pretty awesome. (I warned him ahead of time that a bunch of random Hokies might be snooping around his site).
In the long term, I submitted this data to the redditor who also runs this site. He’s had a busy summer, but the plan is to download all that data for use with the optimized search and tool options on his site.
Long story short, it took me a few years, coinciding with the onset of the pandemic, to obtain, scan, modify and sort individual signer/recruitment data that I found on eBay. I basically figured out how the 247 Composite works, then retroactively applied that format to create a database. In the following sections I will describe the data contained in each worksheet and have added some graphics to show how the data can be sorted and used. (I am posting graphs showing Hokie curves where many are available – links are provided to the other graphs if you wish to browse.)
Fig 1.1: 114,062 Signature Scores plotted in 43 Signature Classes, from highest to lowest
This spreadsheet has by far the most data, but there really aren’t a ton of ways to graph it. All team data comes from the individual signatories data in this spreadsheet.
Fig 2.1: Signatory threshold scores by season
Fig 2.2: Signatory STAR score by season, in percentage
Fig 2.3: Database teams and signatories rated by year
These charts show a blueprint of how data looked before and after the Internet age. Basically, before the internet, you will usually only find ratings on about 1/2 of FBS signers. So with the older data, all unranked players end up having the same score, which is what I started to call the “Floor”. I used this as a threshold for 2-3 star players. Beginning around 2006, most FBS signers began to have some sort of individual score. From this point on, there are far fewer “2-star” signers in the database.
There are still signatories without ratings today; in fact, COVID appears to have temporarily increased the number of players who haven’t been rated well enough to have a composite score. These players simply do not have a score associated with them and they do not affect a team’s score in any way. This affects in particular JMU’s FBS Introductory Course (2022). They have 4 unranked signatories. Even though they have 8 rated transfers, the transfers do not count towards the team’s recruitment scores. Thus, JMU’s 2022 signing class effectively has no class score.
Fig 2.4: Total number of FBS teams and FBS signers per season
Fig 2.5: Average, minimum and maximum number of signatories per class and per season
These graphs show a bit of what college football has looked like over the years and how it has affected this database. I am listing MOST FBS programs, but I have essentially omitted all football programs that do not use scholarship rules and limits. This means that the Ivy League (1980-81) and Service Academies (all seasons) do not show up in my database at all. Service academy signature courses vary wildly, even in the age of the internet they have not publicly released their signature courses. They’re not really competitive for Blue Chip rookies – basically they give full turns to a lot of well-rounded individuals – a lot of them are really good athletes, so their classes are full of respectable “signers”* * and some seasons they I will “sign” some *** players too.
In 1981, the NCAA had a purge of schools that did not meet the requirements to compete in Division 1A. The Southern, Southland, Ivy League, and (eventually) Missouri Valley conferences transitioned to FCS beginning in 1982. In 1992, the scholarship limit for FBS (Div 1A) was increased from 95 to 85 per school. 1989 was the start of this transition, so an average signing class went from 25 signers to 22.5 during this time.
We are currently in a new period of adjustment with the transfer portal (coupled with the 2020 season Covid eligibility queue) so by around 2025 a new average signing class size should be related. (Probably end up somewhere just under 20).
Fig 2.6: GLOBAL signatories by major producing states (VA is on the second graph)
Fig 2.7: BLUE CHIP signatories by main producing states (VA is on the first graph)
While it might seem like GLOBAL Signatories by State is a useful metric to use in determining which states are producing the most talent, the truth is, that’s really not the case. When Richmond and William & Mary descended on FCS in 1984, the number of signatories from the state of Virginia plummeted. It was not a valid indictment against Virginia High School Talent. The recruitment of 2 and 3 star players is simply based on geographical convenience. If the state of North Dakota were to go FBS, there would suddenly be half a dozen signatories from that state each year (which is about half a dozen more on average than there are now ).
Blue Chip production is the metric to use to determine state productivity over time. These players are signed by someone somewhere, regardless of the geographical designation of FBS schools available across the country.
Fig 2.8: P5/G5 distribution, teams and signatories
In 1980 there were a similar number of automatic qualification programs (P5 and independent equivalent) and non-QA programs (G5). Following the 1982 purge, the number of G5 programs dropped dramatically, and over the next four decades essentially grew again.
Over this period, the number of P5 programs has been fairly consistent, hovering around 65.
Fig 3.1: Average team scores from #41-50 (VT is ranked #44 from 1980-2022)
For the team data spreadsheet, I calculate team scores in 3 different ways. First, I basically made an “estimator” of the 247 composite team score. I don’t like this method, so I’m also including the raw average signer score and an “adjusted” average score, which is the method I prefer (and which appears on these charts).
In truth, I’m still deferential to 247 Sports, because this “Adjusted Average” team rating method actually appears in their explanation of team rating:
Fig 3.2: Scoring formula of 247 sports teams
First of all, 247 Sports doesn’t even use this method anymore; they changed their methodology a season after introducing the team score calculator. Basically, this old method only really counted the top third to half of your class. But watch out for that “C”. (They stopped considering that “C” with their new method, BTW.)
Fig 3.3: A 3D image provided by 247 Sports to show their old methodology
Fig 3.4: A 2D representation of how this old method actually affects tabulating a team score
So they settled on a new methodology that would change the team’s score no matter where a new signer was added to a team (obviously a higher-scoring rookie moves the needle a lot more).
Fig 3.5: A 2D approximation of the “New” 247 Sports team score
So back to my favorite – it’s similar to taking the raw average. I just figure out the size of the average signing class each season (lately it’s been 22 or 23 players per class), and I just take the average of the top 22 or 23 players for a signing class.
It’s the “C” from above.
There is a small advantage for larger classes and no penalty with smaller classes. Basically, most programs are pretty good at keeping 85 fellows on a roster. Thus, class size from year to year is, for the most part, unimportant. Teams that consistently have large classes simply more actively manage or shuffle their roster. So they get a little bump for that, but for the most part, average rookie rank is still the best method of determining how programs are doing from year to year.
Fig 3.6: ACC Recruitment Trends
Fig 3.7: Big East/AAFC Recruitment Trends
Fig 3.8: Recruitment trends of intermediate and minor freelancers (VT was a MEDIUM freelancer)
These graphs show the scores of the teams in the different conferences. As a general rule, when conferences change, I’ve shown the current conference in bold and other conference affiliations somehow in dotted or crossed out lines.
When I started this project, I tried to sort all the first FBS freelancers into P5/G5 categories. I found Independence to be more of a spectrum, and used recent schedules and opponents to determine “peers” for Minor (G5), Medium, and Major (P5) schedules. Virginia Tech was an independent average using this method; we were pretty split between playing other G5 and P5 teams (but less than half P5 opponents, or peers). Basically, for these teams, any conference shakeup could go either way. And in truth, he did.
When I developed the team scores, I ended up grouping middle independents with G5 schools (they basically recruit middle, but with no conference affiliation in the TV era, they were always disadvantaged.
In 1997, the last Middle Independent (ECU) joined CUSA, so this category disappeared. After that, either you were Our Lady or you were tiny. Until BYU becomes independent, that is. I haven’t reintroduced this category to the database, but I dissected this for a Reddit thread about a year ago and found that BYU would have essentially been an “average” independent using this same peer methodology /hourly.
EDIT: PS, I made this graphic to show how conference recruiting has changed historically, along with a graphic that shows how I would project new conference rosters to change in the future:
Fig 4.1: Historical and projected trends in conference recruitment