(I’m lazy so most of this text was written by ChatGPT. Apologies if anything is unclear).

Every year, the NFL draft showcases the best college football players in the country, as they transition from the collegiate level to the professional ranks. However, not all colleges are created equal when it comes to producing NFL draft picks. While some schools consistently send large numbers of players to the draft, others struggle to produce any top-tier talent.

But what if we could adjust for recruiting rankings? Would that change the landscape of which colleges are best at producing NFL draft picks? After all, we know recruiting ranking is correlated with the probability of getting drafted, so it is unclear to what extent teams actually develop NFL players vs simply recruit NFL players. In this post, we'll take a closer look at which colleges are the best at producing NFL draft picks, while taking into account their recruiting rankings. By doing so, we hope to shed light on which colleges are truly the best at developing top-tier talent, and which ones may be relying more on incoming talent.

 

The idea

To conduct this analysis we will attempt to build a “null distribution” that tells us how many NFL players we should expect to produce for a given recruiting class. Specifically, we can do this by using the composite ratings of each player in the recruiting class to create a pool of similarly rated players. From this pool, we will randomly sample players to create multiple "null classes" that serve as a comparison group. We will then track the number of NFL draft picks produced by each null class, and use this distribution to determine how many draft picks we would expect a class of similarly rated players to produce by chance alone. By comparing the actual number of draft picks produced by the class in question to this null distribution, we can identify which colleges are best at developing their players.

 

The data

We’ll consider recruiting classes from the Harbaugh era where most players have finished their college careers (i.e. the 2015-2018* classes). Below is a histogram of the composite rating of all players in the CFBDatabase for this time window as well as the empirical probability a player in each bin of the histogram gets drafted:

*Not everyone from the 2018 class is out of eligibility, but most players who are going to be drafted from that class have probably moved on by now so I included it. The results aren’t too different if we only go up to 2017 though.

 

Some general thoughts:

• I’m impressed how well the composite predicts draft probability.
• The top ~25 players in each class (the rightmost bin) have a 64% chance at getting drafted
• The next bin (guys ranked ~25-50) is also pretty high (45%), but it starts to drop off pretty significantly after that
• Not all 3*s are created equal: About 15% of highest 3*s are drafted, while mid 3*s are closer to 7%
• The bin widths I used were 0.015. Any smaller and the draft probability curve gets super wiggly (especially at the top end), so what you should take away is that while a guy rated 0.95 is going to be more likely to pan out than a guy rated 0.9, he might not necessarily be more likely to pan out than a guy rated 0.94. (i.e. rankings matter, but they're noisy)

Evaluating player development

To evaluate the effectiveness of player development at different colleges, for each recruiting class, we will create 1 million null classes as described previously. By tracking the number of NFL draft picks produced by each null class, we will create a distribution that shows how many draft picks we would expect a class of similarly rated players to produce by chance alone. We will then compare the actual number of draft picks produced by the class in question to this distribution, and calculate the percentile rank of the actual class within the null distribution. By using this approach, we can gain a more nuanced understanding of which colleges are truly the best at developing top-tier football talent, and which ones may be relying more on incoming talent than player development. For example, here are the combined results for Michigan’s classes:

So based on recruiting ranking alone, we’d expect Michigan to have 17.8 total players drafted from the 2015-2018 recruiting classes, but in real life Michigan had 30 players drafted, beating 99.96% of null classes.

Results

Finding the percentile rank for each school shows us which schools actually do the best at developing talent:

Some final thoughts:

• Iowa being #1 shouldn't be surprising. They’re obviously not known as a recruiting juggernaut (7.4 expected draft picks), but they’ve been producing a solid number of draft picks (18 actual picks)
• Florida, Michigan, and Penn State rounding out the top 4 checks out, as they’re all schools that recruit well but not elite, yet churn out draft picks (even if Florida doesn't churn out wins)
• Georgia being #5 is seriously impressive given how well they recruit. And it seems like they’re only getting better
• Minnesota, Wisconsin, and OSU are other Big Ten schools that seem to be very good at developing players, but not quite elite
• Michigan State is the worst in the Big Ten and among the worst in the country at developing talent (Expected draftees: 10.9, actual: 6). A far cry from the peak of the Dantonio era
• The Ragin' Cajuns' don't have a particularly staggering draft record (6 picks), but they're nearly tripling expectations (2.3 picks) landing them 7th in this metric
• I was expecting Clemson to score higher than they did
• Texas is the most average school: 18.5 expected draftees vs 18 actual
• Oof Florida State. 22 expected draftees vs only 12. Worse than Army and Navy (who had a combined 0 draft picks)
• USC has also not been good. 24.3 expected vs 18 actual.

 

Disclaimers

• Transfers get claimed by the school they originally went to (e.g. Michigan gets credit for Charbonnet but not Olu)
• Looking at whether or not a player was drafted (vs draft position or even round) is a very coarse metric. The first overall pick counts the same as Mr. Irrelevant 
• Similarly, players can be successful in college but not drafted
• An alternative interpretation could be that the results show which schools are the best at identifying talent that the recruiting industry missed