When the NFL season kicked off week one last weekend, it marked the conclusion of yet another tumultuous offseason. Teams moved off of certain players to bring in replacements, some of whom were signed to big-money deals and some of whom were playing at or just above the veteran minimum salary.
Most notably, this offseason saw the conclusion of the decline in value for the running back position. Several high-profile halfbacks — including Ezekiel Elliott and Dalvin Cook — were released by their former teams, and had to wait months to resign with a new team. Overall, running backs make significantly less than other offensive positions, and have barely seen collective salaries increase much in the last decade.
This decline in pay for running backs is correlated to the amount of perceived value a player is expected to provide during the length of their contract.
Similarly, where running backs are selected in the draft shows how much they are thought to help the team during their career. That, too, tells a story of decline. In the last five NFL drafts, only six running backs were taken in the first round. Compare that to the six drafts from 1992 to 1997, when teams used first-round picks on running backs 19 times.
For Cade Massey, this shouldn't be a surprise. Massey is a practice professor in the Wharton School's department of Operations, Information and Decisions, and one of the pioneers in football analytics. His work focuses on determining how good NFL teams are at evaluating talent in the draft, based on the correlation between the value generated by a player and their draft order. One of the things he's found is that using draft capital on a running back is not necessarily a good idea.
"In the late 2000s, we saw that the only place that you would not expect to generate surplus in the draft was drafting a running back in the top of the first round," Massey said. "Every other position at every other point in the draft, your surplus is positive, because the rookie wage scale keeps things so low, but for that one position at that one place in the draft, it was negative and we had no idea until we ran the curve and so that was our first inkling that that's where running backs were."
Alongside the decline in the running back position has been the increase in the value of having a solid quarterback. Similarly, due to the amount of money top-tier veteran quarterbacks earn — the top 15 highest-paid players by average per year are all quarterbacks — there is an impetus to win while the team still has a young quarterback, which impacts their draft value.
"When we ran our first analysis, quarterbacks hadn't separated as much from the other positions as they have now these days," Massey said. "You really need two values of draft pick curves: one for quarterbacks and one for everybody else. There's still a little bit of debate on that. It's not completely clear. We're doing some more work on it now. But to a first approximation, I think it's fair to say that quarterbacks are so much more valuable than other positions that the advice we'd have on the use of draft picks is different for them than it is for anyone else."
For Massey, not only have the different positional values in football changed, but NFL front offices have become much more numbers driven as well. They have started incorporating calculations similar to his into their analyses.
The use of big data and analytics in the NFL is relatively new, only beginning in earnest in 2014 when then-Philadelphia Eagles coach Doug Pederson sought a more numbers driven way of decision making. Additionally, teams began using player-tracking technology to analyze player tendencies.
"We've got so much more data [than in the past]," he said. "The teams have invested more in their analytics departments. There are more people working on this problem before we've gotten more sophisticated in our models … and then slowly, analytics-oriented people are gaining power within some of these front offices."
With all of this effort being placed in numbers and statistical analysis, have teams gotten better at using draft picks wisely on players who will ultimately produce value? According to Massey, yes, but not by much. From his work, there is a slight uptick in the correlation between draft position and value produced, but not as big of one as would be expected, especially given the increase in investment and sophistication of draft evaluation processes.
Massey attributes this largely to the difficulties inherent in running an NFL front office, which he compared to the stock market.
"It's just tough to beat the market opinion, not because people are dumb, but because a lot of smart people are playing the same game," he said.
However, there are also factors unique to the NFL environment. Among them is a level of "irreducible uncertainty" that Massey sees in player evaluation. That could change in the future with new statistical tools, but for now, Massey says that there's a ceiling for predictions.
Another issue that complicates the work of NFL front offices — and the researchers that study them — is the difficulty in measuring a player's total value. The complexity of the NFL, and the dependence of each player on all the others, makes determining each player's impact on winning and losing difficult.
"This is one of the real challenges in football [and] in many [other] sports, but it's so far from being solved," Massey said. "We're going to be working on it for a long time and we'll never know it for sure, we'll never know it perfectly. That's one of the things that's interesting. We know that these outcomes are interdependent, we know that people work with other people on the field to make these things happen, but it's very hard to parse exactly what that point is. It's very hard to understand what the interactions are."
Despite these challenges, Massey and his peers are continuing on in their work of using numbers to solve the mysteries of the NFL.