Better combine numbers don't necessarily mean success for linemen
This story appears in ESPN The Magazine's March 3 Analytics Issue. Subscribe today!Four days before Jameis Winston hoisted the Heisman, John Urschel struck a more thoughtful pose in New York. The two-time All-Big Ten guard accepted the William V. Campbell Trophy (aka the academic Heisman), capping a Penn State career that included a bachelor's and master's in math and a second master's in math education that is nearly complete (all A's). If Urschel wasn't pushing the sled, he was building his CV. In his final season, he not only earned third-team AP All-American honors but also taught Math 232 (integral vector calculus), and he recently proved the Urschel-Zikatanov Generalized Bisection Theorem in a paper titled "Spectral Bisection of Graphs and Connectedness." It's been accepted by Linear Algebra and Its Applications, the top journal in spectral graph theory. "This may not be cool to your readers," says the 22-year-old, "but to a math-savvy person, proving your own theorem is pretty sick." Yes, an offensive lineman actually said that.
WHILE PREPARING FOR the NFL combine on Feb. 22, I worked hundreds of hours on my 40-yard dash, 225-pound bench press, 20-yard shuttle and other drills, hoping that better results will deem me worthy of a draft pick. But as I was running and lifting, I couldn't help but wonder, from a purely analytical standpoint, if those skills even relate to my future success.
So like any good mathematician vexed by a question, I searched for an answer. Then The Mag approached me to brainstorm ideas for its Analytics Issue. We decided I should look at combine results of the 400 offensive linemen invited and graded by Scouts Inc. since 2006 to determine what, if anything, strongly indicates success. Using ESPN Stats & Information's data -- including body measurements, combine results, draft position and NFL performance (e.g., starts) -- I treated three problems: 1) how best to predict a lineman's draft position, 2) that prospect's success in terms of NFL starts, and 3) whether a fringe prospect will be selected. My results are below.
The projection of how a college player's performance will translate to the NFL is far from an exact science. As my fellow mathematicians might say: There is agreement on the sigma-algebra but much debate on the proper measure. [Ed.'s note: Just Google it.] This is especially true for evaluating O-linemen. That's because I can count on one hand the times I blocked 40 yards downfield. And believe me, benching 225 doesn't compare to blocking Dolphins defensive end Jared Odrick or Bengals defensive tackle Devon Still, both PSU alums whom I faced day after day in practice.
I contend that the offensive line is the only spot in football where intangibles like toughness and determination are as key as athleticism, a theory backed by my study. For instance, when I tried to predict where a lineman would be drafted, all combine results -- including body measurements -- lacked significant correlation. The one factor that proved most helpful was a lineman's position. On average, tackles tend to go nearly a round earlier than guards and centers because of the increased importance of the edge pass rusher in the NFL.
So if measurements and combine tests don't account for draft position, what are the relevant factors? The answer that naturally comes to mind is on-field performance, but mathematically speaking, the tale of the tape doesn't prove as accurate as one might think. When I used Scouts Inc.'s prospect grades to measure college performance, it was reliable only in early rounds. Considering only the linemen selected in the first 150 picks, the grades had a correlation of 81 percent to overall draft position. For the remaining picks, the correlation plummeted to 17.5 percent, because although scouts agree on what makes a great lineman, there is still much debate on what makes a good lineman.
Spot an every-down starter
ESPN: The Mag
Ty Wenger delves into the newest edition of ESPN The Magazine to look at the ascension of Paul George & the beautiful mind of an NFL hopeful with Pablo S. Torre & LaRue Cook.
When a lineman is taken in the top 150 picks, his draft spot has a mediocre 41.2 percent correlation to his average starts. For a lineman outside the top 150, the correlation crashes to 11 percent, proving that across all rounds and especially in the later ones, teams haven't found absolute measures for predicting a surefire starter. A scout can form an opinion based on one in-person meeting, while his GM can have very different opinions about what traits a long-term NFL starter should possess. These subjective differences are what make the NFL draft more crapshoot than science.
Find a lineman in the rough
I entered the combine ranked the No. 7 guard by Scouts Inc., so I pray that I possess these intangibles that NFL scouts (and analytics) can't fully capture, eventually proving that I was overlooked, just as I went from an unranked recruit to an All-American.
But my study was not a complete wash. I did have success determining the likelihood of a fringe prospect like me to be drafted. Two combine tests that proved statistically significant in doing this were the 40-yard dash and 20-yard shuttle. Focusing on the times of linemen taken outside the top 150 and those who weren't drafted, I produced a model that is a good predictor of being selected, with a concordance statistic (or c-stat) of nearly 75 percent. (The lowest possible c-stat is 50 percent, and any c-stat approaching 80 percent is highly reliable.)
The results show just how important the 20 and 40 are to scouts and NFL teams when looking for a lineman in the later rounds. Of the 161 players in my data set, 27 ran the 20 in under 4.6 seconds, and only six went undrafted, likely because all six ran the 40 in 5.15 or slower. Similarly, 24 ran the 40 in under 5.15, and only four went undrafted, likely because all four ran the 20 in 4.62 or slower.
So my advice to fellow under-the-radar prospects: Improve your coverage of linear and lateral area. Or in lineman terms: Get your ass in gear.