I’m an archaeologist. And I love college football.
In particular, I enjoy following college football recruiting. I think this is because, as an archaeologist, I delve backwards in time to try to find big-picture trends in order to understand why things are the way they are now. So, for example, I study the context for the origins of agriculture to help understand the positive and negative issues we face with our current diet.
Similarly, tracking trends in college football recruiting helps us to see some of the long-term trends that ultimately translate into wins and losses on the football field. Diving into the mountains of data freely available on recruiting websites like Scout, Rivals, ESPN, and 24/7 Sports give us a way to dive into the past like an archaeologist excavating an ancient village.
This leads me to my favorite all-time question on college football message boards: “Do recruiting rankings matter?” In other words, to what degree does the quality of recruits translate to actual wins and losses? I think I’ve seen just about every variant of the arguments on either side.
First, there’s conjecture as to how accurate recruit rankings are. The major recruiting websites generally assign a ranking from one to five stars. More stars equal a better recruit. The 24/7 Sports analysts have created a composite rank based on averages from multiple services and is arguably the current gold standard.
As for the debate on whether highly ranked recruiting class matters, there are two basic positions. On one side you have the “star-gazers” who believe these rankings accurately predict how well a team is likely to do in the future. These individuals generally point to the fact that the teams with the best records generally get the best recruits and this creates a positive feedback loop. The phrase that gets frequently used in this context is that “It’s not always about x’s and o’s, it’s about jimmies and joes.”
Others argue that the rankings assigned by recruiting services are unreliable at best (or a scam at worst). They will frequently cite examples such as Boise State (who perennially over-achieves despite low recruit rankings) and Texas (who regularly under-achieves). Many in this camp instead argue that the best way to gauge recruits is to see which, and how many, programs have offered scholarships. For example, a recruit with only offers from Western Kentucky and North Texas is likely not as good as an athlete who has offers from LSU, Alabama, Ohio State and 10 other historically top-notch programs. In other words, successful coaches at big-time schools have a better eye for talent than the analysts working for the recruiting services.
Why does this debate exist?
As a social scientist, I don’t find the debate surprising. In archaeology, and the social sciences more broadly, we stumble across patterns that are (for lack of better words) fuzzy. College football recruiting is no different.
To illustrate this, I downloaded the Sagarin Ratings, which ranks programs based on wins and losses in a way that accounts for strength of schedule. A higher rating translates to a higher rank.[1] I also downloaded the 24/7 Sport Team Talent Composite, which ranks teams at the beginning of the season based on the quality of the recruits they have on their roster.[2]
Figure 1. Sagarin Rating vs. 24/7 Team Talent Composite
By creating a basic scatterplot and linear regression line through the points, a couple things become pretty clear. First, we see relatively clear (albeit fuzzy) positive correlation between success in recruiting translating to success in wins and losses in the subsequent season. Another way of interpreting that graph is that more than half of the variation in wins and losses can be attributed to recruiting.
However, second, we see plenty of exceptions to the rule. Or, in other words, there are plenty of examples of teams who finished well above, and well below, where they should have based on the talent on their roster at the start of the season. Part of this is likely due to random chance, or luck. Alternatively, I would argue that the distance from the lines gives us a way to demonstrate which teams truly over- and under-achieved based on the talent they had on hand at the start of the season.
Figure 2 – The Top 20 Overachieves and Top 20 Underachievers.
In the top 20 column, there’s a list of the darlings of the 2017 season, headlined by the best season by Army in over 20 years, two of the four teams in the college football playoff (Oklahoma and Clemson), teams with well-known gurus for coaches (TCU, Penn State, Washington and Washington State), solid programs that perennially seem to do more with less (Wisconsin, Iowa, Northwestern, and Kansas State) and teams with coaches who had their coaches hired away (UCF) or coughed up new contract extensions (Iowa State, Purdue, Memphis).
On the other side, a murderer’s row of teams who fired their coaches, are rebuilding their programs, or just recently made the jump to the FBS from the FCS. There are some truly remarkable teams on that list, headlined by Tennessee, Florida, Arkansas, and Florida State – programs that are perennially strong, yet had atrocious seasons that led to coaching changes (all were fired except for Jimbo Fisher, who jumped from Florida State to Texas A&M).
What does all of this tell us about using recruiting rankings to understand “the bigger picture” of college football? It tells us that recruiting matters a lot, and explains more than half of the variation in wins and losses. Yet, that other half that isn’t explained by recruiting is likely a combination of luck (both good and bad) and coaching.
In other words, there is a clear trend, yet some pretty obvious exceptions to the rule.
This is the root of the yearly debate over recruiting rankings.
[1] I used the “rating” instead of the “ranking” because the rating is a continuous variable, and is more appropriate to use in a basic linear regression model when extracting and analyzing the residual values. The Sagarin Ratings were downloaded on Dec. 10th, before the start of the bowl season.
[2] I used the total points rather than ranking because it a continuous variable.
