Taking a Deep Dive into the NFL Draft

Optimistic Browns fans watch their favorite team play.

The NFL Draft is one of my favorite events of the year; for every team (yes, even the Browns!) the Draft represents the chance to turn your fortunes around. While much of the focus on college athletics focuses on the negative, there is something undeniably captivating about watching people realize their dreams (as just one example, I could watch James Conner get drafted by my team, the Pittsburgh Steelers on an endless loop). However, as should be obvious by this point, I like data and wanted to spend some time digging into some nuances of the NFL Draft.

The first step in my process was to actually find some data to utilize. Personally, I love the entire family of sports reference sites but given the nature of my sports interests, I most regularly use Pro Football Reference and Basketball Reference. Instead of manually collecting data, I wanted to see if it was possible to scrape the data (for the uninitiated, scraping is a coded, automated means of pulling down data that can be performed in a variety of programs; in this case, I used R). Luckily for me, here I didn’t have to reinvent the wheel as fellow MIT student Steven Morse provides an incredibly helpful tutorial that I was able to use as a guide to pull draft data from 2000 to 2016.

(h/t to Mark Stevenson for the data science meme above)

In contrast to basketball where side by side player evaluation is easier based on any number of metrics, comparisons across position in the NFL are considerably more difficult, particularly when looking at both offensive and defensive players. However, Doug Drinen once set to encapsulate the contributions of all players in a single metric, which he called “Approximate Value” (described in much more detail here) and weighted Career Approximate Value, will be the relevant metric I will be using henceforth.

Finally, the last bit of explanation necessary before formally journeying through the data is some additional context from brilliant football analyst and 2017 Sloan Sports Analytics Conference alum, Chase Stuart. Obviously, all draft picks are not created equal; the production that an NFL team should reasonably expect from the first overall pick should be much higher than a player drafted in the 6th round. Chase created a value chart showing the relative expected production differences here but I would encourage any data-inclined football fan to check out his work on Football Perspective (for the extremely curious, he also later did a subsequent update to this draft pick analysis but in the interest of simplicity, I stuck with his original work).

Whew! That had more exposition than a bad movie but if you are still with me, let’s look at some charts.

Aside from the pretty colors, a couple of things stand out immediately from this chart. The most evident thing to me is the outlier in the upper right quadrant, a player who was drafted 199th overall but has created the most value for his team of any player over the time period in question. Even the most casual of football fans can likely name Tom Brady, 5x Super Bowl champion and a player who is generally regarded as the Greatest of All Time (in sports slang, the GOAT).

The GOAT with a goat.

Drew Brees, another QB who has put up gaudy stats over the course of his career only falls slightly below Brady. The other two players that stand out unfortunately fall on the opposite end of the spectrum as they are the only two skill position players to produce negative Approximate Values. One of them is Ryan Lindley, a late round quarterback drafted by the Arizona Cardinals who through an unfortunate set of injuries (both starter Carson Palmer and second-string QB Drew Stanton were out) had to play in a playoff game that went predictably poorly (16–28, 86 yards passing, 1 touchdown, 2 interceptions, and 4 sacks).

The other is 2016 number 1 overall pick Jared Goff, who was acquired at a considerable cost in terms of draft capital after Los Angeles traded up. In the history of the NFL, including Goff only 9 rookie quarterbacks who played at least 4 games had negative AV in their first season in the league. However, all hope is not lost yet for Rams fans; of that list, 2 QBs like Goff also went first overall. These two, Eli Manning and Alex Smith, have proven to be at least competent quarterbacks in their career.

Beyond these outliers, many features of the scatterplot seem to make intuitive sense. There is an inverse relationship between career Approximate Value and when a player is picked as one would expect. Quarterbacks are generally taken highly in the draft, in particular, as the first overall pick, which reflects how the position is valued in the modern NFL. Finally, while the NFL markets all of its stars, on a pure numbers basis, a vast majority contribute very little to their team.

Now, let’s take a look at the chart, which utilizes Chase Stuart’s Expected Draft Pick Value rather than the pick itself outright.

Using Expected Draft Pick Value rather than pick number helps to condense the scatterplot considerably as the highest value asset (the number 1 pick) has a capped value of 73. This model also reinforces the value of hitting on a late round pick like Tom Brady, whose pick cost is 1/10 of the #1 overall pick. While again the colors of the chart are certainly attractive, on an individual player level or team level, it’s hard to discern any meaningful patterns.

Note: Given that this weighted career Approximate Value calculation is meant to look at players over the course of their entire careers rather than an individual season or two, I’ve made some adjustments to the Expected Draft Pick Values of anyone that has played less than 5 seasons (hence why we don’t see Jared Goff’s negative point where it would be expected).

One question that can be answered is which players have provided the most value relative to their draft position. That chart is below:

While it’s nice to have a nice top 10, it was an unexpected result as the data was filtered for players who produced in excess of a first round pick in their careers. Again, Brady is king even using weighted career Approximate Value; to the Patriots, he has been worth more than two additional first round picks! Out of this list of ten, I would expect that 8 are surefire Hall of Famers once they retire, and Jahri Evans and Lance Briggs who went to 6 and 7 Pro Bowls respectively likely will merit some consideration as well.

Looking at the other end of the spectrum, we can also see the players who were the worst values to their respective teams. Unsurprisingly, JaMarcus Russell tops the list. Apparently lighting up my alma mater’s secondary in the 2007 Sugar Bowl is not indicative of the ability to succeed in the NFL.

Finally, I wanted to take a brief look at my own favorite team, the Pittsburgh Steelers.

The Steelers have obviously done quite well in terms of drafting, particularly with some of their later round picks like Antonio Brown. There is a widespread belief that the Steelers are extremely skilled in picking WR talent. However, when narrowed down to this particular position as shown in the chart below, it’s not immediately clear that they’ve been especially skillful in their picks. In the coming weeks, I’ll hope to analyze this result further.

For me, it’s time to return to the lab:

I always was a sucker for this show; until next week, same bat time, same bat channel.
Show your support

Clapping shows how much you appreciated John Martell’s story.