Stat Analysis
Advanced analytics on player and team performance

BackCAST 2021

Clemson Tigers RB Travis Etienne
Photo: USA Today Sports Images

Anyone who has played fantasy football knows that a quality running back can quickly emerge out of nowhere. Although many of these players later fade back into obscurity, several have staying power. Terrell Davis, for example, went from a sixth-round pick to two Super Bowls and the Hall of Fame.

Could this year's draft contain a late-round gem? The odds are always against it. However, BackCAST, Football Outsiders' system for projecting running back success, believes that this year's draft has an unusual number of promising late-round prospects who could surprise.

BackCAST projects NFL running back success based on statistics that have correlated with success in the past. Historically, a college running back who has a good size-speed combination, gained a high average yards per carry, and represented a large percentage of his college team's running attack is more likely to succeed at the NFL level. BackCAST considers these factors and projects the degree to which the running back will exceed the NFL production of an "average" drafted running back during his first five years in the NFL. For example, a running back with a +50% BackCAST is projected to gain 50% more yards than the "average" drafted running back.

BackCAST also includes "RecIndex," which measures whether the player is likely to be a ground-and-pound two-down back, a player who catches passes out of the backfield more often than he takes handoffs, or something in between. In short, RecIndex measures the likelihood that the player records a disproportionately high or low number of receiving yards versus his rushing yards. The two factors significant in predicting RecIndex are receiving yards per game in college and weight, as smaller players are likely to be receiving backs.

BackCAST is based on a statistical analysis of all of the Division I halfbacks drafted in the years 1998 to 2019 and measures the following:

  • The prospect's weight at the NFL combine.
  • The prospect's 40-yard dash at the NFL combine. If he did not run at the combine, BackCAST uses his pro day time.
  • The prospect's yards per attempt with an adjustment for players who had fewer career carries than an average drafted running back.
  • The prospect's "AOEPS," which measures how much, on average, the prospect's team used him in the running game during his career relative to the usage of an average drafted running back during the same year of eligibility.
  • The prospect's receiving yards per game in his college career.

What follows are some of the most notable BackCAST projections for the running back prospects available in the 2021 NFL draft.

Travis Etienne, Clemson Tigers
BackCAST Score: +124.2%
RecIndex: +0.33
Similar Historical Prospects: Dalvin Cook, Reggie Bush

Travis Etienne is one of the top-rated running backs in this draft, so it is no surprise to see him top this year's projections. Although his projection falls short of some of the all-time greats, he is still an excellent prospect who is worthy of his draft status.

Etienne's best statistic is his yards per attempt. Etienne averaged 7.22 yards per carry over his college career, which is outstanding. Indeed, only three running backs in BackCAST's database had a better adjusted yards per attempt than Etienne. The metric definitely has some diminishing returns, as the three top running backs in the metric (Melvin Gordon, Darrell Henderson, and Rashaad Penny) have had mixed success. However, more efficient college running backs have succeeded at a greater rate than less efficient ones, so Etienne's success on a per-play basis is a definite plus.

Etienne's size-speed ratio also better than average. The average drafted running back is just over 215 pounds with a 4.54s 40-yard dash. Etienne weighed in at 215 pounds but was fast, running the 40-yard dash in 4.42 seconds.

Etienne's biggest "weakness" is that he did not dominate his backfield like other top running backs prospects in past drafts. Etienne had an AOEPS of +1.8%, which is better than average but not spectacular. Although Etienne was very efficient, Clemson gave several carries to other running backs such as Lyn-J Dixon, Chez Mellusi, Adam Choice, and others during Etienne's career.

Etienne's final plus is that he was strong in the receiving game. Even if Etienne ultimately does not pan out as a pure runner, he will likely add value as a receiver.

Najee Harris, Alabama Crimson Tide
BackCAST Score: +72.2%
RecIndex: -0.02
Similar Historical Prospects: Marshawn Lynch, LaMont Jordan

Najee Harris may be hard to project because Alabama's unusually talented roster could undermine some of BacKCAST's fundamental assumptions. Specifically, AOEPS assumes that most college running backs, even those that are potential draftees, will not ultimately be viable NFL running backs. Accordingly, if a team is constantly splitting carries with another running back on his team, it is typically a red flag. Most college coaches know their players well and will funnel most of the team's carries to a running back who is talented enough to be an NFL star.

Alabama, however, is the rare case where that assumption may not be true. Harris, for example, had to compete with future first-round pick Josh Jacobs and future third-round pick Damien Harris early in his career for carries. Accordingly, Harris' AOEPS of +2.1% may undersell his true ability.

Further obfuscating Harris' projection is the fact that Harris did not run the 40-yard dash in pre-draft workouts due to injury. Harris does, however, have good size at 230 pounds. A first-round level prospect typically has a 40-yard dash in the low 4.5s, which is where BackCAST has Harris.

That said, despite all of the questions surrounding Harris' projection, it still comes in as a solid 72.2%, which is firmly average for a first-round pick. Accordingly, scouts who like Harris' tape do not necessarily need to shy away from selecting him fairly high.

Chuba Hubbard, Oklahoma State Cowboys
BackCAST Score: +52.9%
RecIndex: +0.11
Similar Historical Prospects: DeMarco Murray, Laurence Maroney

This is the point in the draft where BackCAST departs from conventional wisdom and never comes back. Although most rate Hubbard at around a third-round pick, BackCAST likes him more than any backs other than Etienne and Harris.

Hubbard is not outstanding in any one quality but is solid all around. First, Hubbard's AOEPS is good. Hubbard had a larger percentage of carries than most drafted running backs and outpaced both Etienne and Harris in the statistic. Hubbard's 5.9 yards per attempt is not going to break any records but is (again) better than average. Similarly, Hubbard is a bit lighter than the 215-pound average at 210 pounds, but he is also faster than average with his 4.48s 40-yard dash. While both size and speed are important, history suggests that it is a bit better for a running back to be fast than big, so that tradeoff works in Hubbard's favor. Hubbard is also a little more productive than average in the receiving game, which also brings some value.

Hubbard could be a fantastic option for a team that needs a running back but does not want to spend a premium pick.

Jermar Jefferson, Oregon State Beavers
BackCAST Score: +36.3%
RecIndex: +0.01
Similar Historical Prospects: Cedric Benson, Kareem Hunt

Jermar Jefferson sneaks into the top echelon of BackCAST's projections for one reason only: he dominated Oregon State's backfield very early in his career. Jefferson carried the football 239 times as a freshman, was actually second in carries behind Artavis Pierce as a sophomore, and then reestablished dominance in six of Oregon State's seven games during his junior year. As a result, Jefferson has an AOEPS of +20.0%, which is even a little better than recent all-world prospect Saquon Barkley. Jefferson was also efficient. His yards per attempt improved every year, and he finished with 5.68 yards per carry for his college career.

However, Jefferson's weakness is his raw speed. Jefferson ran his 40-yard dash in 4.6 seconds, which is disappointing for a player who is only 206 pounds. Although his dash time is not a good sign, it is something that many prospects have overcome.

Although the production and athleticism trends each point in the opposite direction, when considered together, Jefferson comes out as better than average. Teams seem to be sleeping on Jefferson, but he is definitely a prospect who could surprise.

Kenneth Gainwell, Memphis Tigers
BackCAST Score: +36.1%
RecIndex: +1.09
Similar Historical Prospects: C.J. Spiller, Javorius Allen

Gainwell has great prospects as a receiving back. Gainwell averaged over 36 receiving yards per game, which tops all other running back prospects invited to the combine. Even though Gainwell is on the small side at 201 pounds, he also has some speed, running a 4.47s 40-yard dash. Gainwell also averaged a nice 5.92 yards per attempt when he ran the ball.

Gainwell comes out with a slightly negative AOEPS, but that could be due to some unusual circumstances. Gainwell had few carries his freshman year but dominated his backfield as a sophomore. Gainwell then opted out of his junior year due to COVID concerns. One could easily argue that, absent COVID, Gainwell would have dominated Memphis' backfield as a junior, which would shore up the one weakness in his BackCAST projection.

Gainwell has a lot of positive signs in his projection, making him a quality mid-round prospect in a draft with a lot of promising second- and third-day options.

Elijah Mitchell, Louisiana Ragin' Cajuns
BackCAST Score: +31.9%
RecIndex: +0.18
Similar Historical Prospects: Jordan Todman, Latavius Murray

During his tenure at Louisiana, Elijah Mitchell was out-carried by teammate and fellow draft prospect Trey Ragas. A split backfield is usually a recipe for poor BackCAST projections. However, Mitchell reportedly dropped several pounds before his pro day and ran a sub-4.4s 40-yard dash, which gives his projection a significant boost. In addition to his speed, Mitchell was efficient at Louisiana, averaging 6.2 yards per carry, slightly more than Ragas. Mitchell's 40-yard dash and his good yards per carry are enough to overcome his relatively low carry numbers at Louisiana, and as a result, BackCAST grades him as an above-average prospect who could be well worth a mid-round pick.

Stevie Scott III, Indiana Hoosiers
BackCAST Score: +27.4%
RecIndex: -0.10
Similar Historical Prospects: Ladell Betts, Correll Buckhalter

There is no way around it: Scott's 4.5 yards per attempt is poor compared to the average drafted running back. However, it's also true that Indiana struggled mightily to run the ball during Scott's entire career. Indeed, the Hoosiers averaged only a paltry 3.63 yards per carry when Scott was a sophomore and an even worse 3.13 yards per carry in Scott's shortened 2021 season. Scott was definitely not in an ideal situation and it suggests something that, when the Hoosiers did run the ball, they very often gave it to Scott. Scott has good size at 225 pounds but ran a slow 4.66s 40-yard dash. Although not a perfect prospect, Scott could nonetheless be an intriguing pick for a team looking for a big back.

Rhamondre Stevenson, Oklahoma Sooners
BackCAST Score: +14.2%
RecIndex: -0.02
Similar Historical Prospects: Alfred Morris, Montee Ball

Rhamondre Stevenson will command a relatively high pick, but at least from a BackCAST perspective, there is not much to distinguish him from other prospects who will be available later in the draft. Stevenson had great yards per attempt (7.2), but he had only 165 career carries. BackCAST assumes that a high-efficiency, low-carry back like Stevenson would regress to the mean, so his 7.2 yards per attempt on 165 carries translates into an adjusted yards per attempt of around 6.0 yards per carry—still good, but not quite elite.

Moreover, after Stevenson arrived at Oklahoma, he never dominated the backfield. As a senior, he outrushed fellow back T.J. Pledger by only six attempts (although to be fair, Stevenson also missed some time). Stevenson has good size but is relatively slow, running a 4.64s 40-yard dash.

Overall, Stevenson grades out as an above-average prospect, but only slightly so.

Michael Carter, North Carolina Tar Heels
BackCAST Score: +10.1%
RecIndex: +0.21
Similar Historical Prospects: Sony Michel, Ryan Williams

Javonte Williams, North Carolina Tar Heels
BackCAST Score: +5.9%
RecIndex: +0.15
Similar Historical Prospects: Javon Ringer, Cadillac Williams

BackCAST is typically skeptical when two running back prospects from the same college team have high projected draft positions. Many of these "dynamic duos" have underperformed in the past. Take Reggie Bush and LenDale White, who hailed from a dominant USC team, or Ronnie Brown and Carnell Williams, who were the second and fifth overall picks in the 2005 draft, respectively. Although none of the four were out-and-out busts, none of those backs lived up to their lofty draft position. Carter and Williams are not quite on the level as those tandems, but the same principle applies. When neither back is able to clearly beat out the other for carries, it is rarely because both are quality NFL talents. It is more likely a sign that neither is an elite back.

Carter edges out Williams because he averaged a very good 6.6 yards per attempt to Williams' good-but-not-quite-as-good 6.3 yards per attempt. Neither player has a great size-speed combination. Carter ran 4.54s 40-yard dash, which is exactly average for a drafted running back, but at 201 pounds, he is significantly lighter than average. Williams is bigger at 212 pounds, but also slower, running the 40-yard dash in 4.57 seconds.

To their credit, Carter and Williams each have potential in the receiving game because both players were both productive catching the football and are both on the light side, which historically has indicated that a player might be more useful running patterns out of the backfield than being a traditional ground-and-pound back.

It would not be unusual for Carter and Williams to beat their projections and succeed. After all, they are both projected to be "above average" for drafted running backs. However, with so much value available in the mid- to late rounds, teams may want to consider using their second- and third-round picks on other positions and finding their running back later.

The following table lists the BackCAST and RecIndex projections of all of the 2021 running backs who received invites to the NFL combine.

BackCAST Projections, 2021
Name School Weight Forty AOEPS Adj Y/A RecYds/G BackCAST RecIndex
Travis Etienne CLEM 215 4.42 1.8% 7.22 21.0 124.2% 0.33
Najee Harris ALA 230 4.52 2.1% 6.02 15.3 72.2% -0.02
Chuba Hubbard OKST 210 4.48 8.6% 5.91 14.5 52.9% 0.11
Jermar Jefferson ORST 206 4.60 20.0% 5.68 11.1 36.3% 0.01
Kenneth Gainwell MEM 201 4.47 -3.0% 5.92 36.8 36.1% 1.09
Elijah Mitchell ULL 201 4.35 -5.4% 6.20 14.2 31.9% 0.18
Stevie Scott III IND 225 4.66 24.0% 4.52 12.4 27.4% -0.10
Jaret Patterson BUFF 195 4.59 16.0% 6.11 8.5 16.0% 0.00
Rhamondre Stevenson OKLA 231 4.64 -6.6% 5.93 15.7 14.2% -0.02
Michael Carter UNC 201 4.54 -4.0% 6.61 14.9 10.1% 0.21
Javonte Williams UNC 212 4.57 -2.7% 6.01 15.9 5.9% 0.15
Javian Hawkins LOU 183 4.46 14.7% 5.78 8.8 3.8% 0.11
Pooka Williams Jr. KAN 175 4.37 7.4% 5.66 20.5 -0.9% 0.65
Trey Ragas ULL 218 4.63 -1.6% 5.99 8.7 -4.4% -0.19
Brenden Knox MRSH 215 4.67 13.0% 5.19 8.2 -9.8% -0.18
Trey Sermon OSU 215 4.61 -9.0% 6.34 10.8 -11.2% -0.08
Kylin Hill MSST 214 4.57 -8.9% 5.58 15.8 -28.3% 0.13
Rakeem Boyd ARK 213 4.64 -1.5% 5.54 11.9 -31.9% -0.02
Khalil Herbert VT 210 4.50 -10.9% 5.78 7.1 -32.1% -0.19
Larry Rountree III MIZ 211 4.70 10.4% 4.99 6.0 -46.6% -0.24
Chris Evans MICH 211 4.52 -15.4% 5.51 11.4 -53.4% -0.02
Gary Brightwell ARIZ 218 4.62 -8.6% 5.34 4.5 -59.6% -0.36
Demetric Felton UCLA 189 4.51 -15.8% 5.07 30.9 -87.0% 0.95

An edited version of this article originally appeared on ESPN+.


13 comments, Last at 28 Apr 2021, 1:59am

3 Two backs, one team

The younger Javonte will be drafted before Carter just like the younger Josh Jacobs was Damien Harris and they're decent (even if Jacobs was over drafted). But like them, they had to share the ball unlike the guys ahead of them. Doesn't necessarily mean a bad thing, just different roles.

Also Felton is incredibly unathletic despite the hype. 

4 65 comments on QBase 18…

65 comments on QBase

18 comments on Playmaker

12 comments on SackSEER

3 comments (before this one) on BackCast, only 1 of which was substantive

The FO community seems to have bought into the "don't draft RBs" mantra to the point that it doesn't even want to discuss who shouldn't draft whom.

9 Have you looked into any…

Have you looked into any fancier rushing efficiency stats, with some kind of situational adjustment or even just some credit for TDs?

I think that Carter's YPC edge over Williams is just due to different usage, with fewer goal-line and short yardage carries weighing down his average. Excluding red zone carries, Williams actually has a slightly higher YPC than Carter, 7.39 vs. 7.34.

10 Differentiating YPC by red…

Differentiating YPC by red zone versus non-red zone sounds interesting.  No idea if it would improve the model.  3rd down carries and TDs might be other ways to distinguish backs who were used a lot in short down-and-distance situations, which would presumably drag down their overall YPC.

11 separating types of carries

Presumably, scouts are doing this. If a person could find success rate for these backs, that would also help distinguish backs who might have shorter YPC from less skill from those with short YPC b/c of usage. As far as red-zone vs. non-red-zone, I think I might use the 10 yard line as the divider. An imperfect proxy would be for any carry in a "goal-to-go" situation, if that is easier for a search tool to separate out than the starting yard line.

Of course, in an alternate universe in which I am a NFL team employee tasked with charting carries for college RB's that my team might want to draft, I might throw cap any carry at 30 yards (speed will show up in other data), give extra credit for carries where a stiff-arm/broken tackle/juke allowed him to gain extra yards, etc. To me, the traits I would be looking for are: ability to overcome sub-standard blocking; ability to hit the correct hole; ability to set up blocks/use them correctly; and ball security. Speed and strength obviously have to meet a functional baseline; work ethic is a must also. The best RB's will jump off the screen at you. No one debates whether Saquon was worth drafting, it was simply how high. I'm talking about evaluating who might be the best out of 5 RB's that might go between the 3rd and 4th rounds, or something like that.

12 A simple thing you could do…

A simple thing you could do just with box score stats is the Adjusted Yards Per Attempt approach, where each touchdown gives a bonus of say 20 yards (that's what PFR does for passing stats). So 10 carries for 80 yards & no touchdowns would be worth the same as 10 carries for 60 yards and a touchdown - 8.0 adjusted yards per attempt. Williams is ahead of Carter in career adjusted yards per attempt as long as a TD is worth at least 9.5 yards.

If you go farther down the road of adjusting for situation, and you have play-by-play data, then maybe you just go all the way to something like VOA or EPA/attempt.