Projected Running Back Success Model: An Analytical Introduction
I must admit that writing this is a little nerve-wracking. I've always been of the mindset that fantasy football was luck. And sure, there are a lot of luck-based elements to the game; most of which revolve around injuries or stupid seven-days-before-the-start-of-the-f*cking-season running back signings (I hate Bruce Arians). But what if I told you that I've attempted to create a model that will help you evaluate early-career running backs at a new level? What if there are metrics that actually have a high degree of significance in the fantasy success of a running back?
What if none (or at least not many) of these metrics have been analyzed - or at least published - for you to have at your disposal when entering rookie drafts throughout each offseason? I'm here to fix that.
The notion that I could create my own predictive metric came from some research into metrics that are predictive of wide receiver success. Bryan Edwards kept coming up in Fantasy Twitter as a "can't miss" prospect because of his breakout age and college dominator score. I cross-referenced with my film grade of Edwards and realized I had come to a similar conclusion, though through a different lens. My film grade of Edwards checked out as higher than consensus in most places, and I was extremely fond of him as a prospect. So I dove into the predictive analytics of wide receivers. And then I searched for predictive running back success and only found something that said "take last year's points and subtract touchdowns." I then came across Blake Hampton's model over at The Undroppables. His model is extremely well done, and he's since moved onto FTN Network. I'm not discounting what he put together (at all!), but I wanted to narrow the level of variance down.
My model focuses less on projected fantasy points per game, but rather on the odds above average that a rookie running back will have a top-24 fantasy season in his first two seasons. Let's look at how the data was compiled.
Photo by Kevin C. Cox/Getty Images
The first thing I wanted to see was: what do all top-24 running backs have in common? And the answer was complicated. When guys who would seemingly be considered outliers consistently get into the top-24, it's hard to identify what the great backs have in common. A good example of this is Phillip Lindsay vs. Christian McCaffrey. One was a first-round pick; one was undrafted. Both have reached the top-24 in each of the last two years; thus, making it difficult to identify correlation.
So because it was difficult to identify commonality, I decided to look at the five-year averages of some college metrics as it pertained to top-24 backs. The data points I initially analyzed were:
College Target Share
Here's what the top-24 from 2019 looked like:
You're probably wondering what the color designations mean. Green means they fell above the five-year average threshold, while red means they failed to meet that threshold.
Here are the five-year averages for each of the metrics:
College Dominator - 32.35%
College YPC - 6.02
College Target Share - 9.87%
SPARQ-x - 115.68
40YD - 4.53s
Speed - 103.86
Burst - 119.40
Agility - 11.30s
Bench - 19 reps
After I created the thresholds, I decided to figure out what actually matters. That's where correlation coefficients come in. In traditional statistics, 0.7 is a good correlation coefficient. What I have found, across the board, in fantasy football is that there's not much that correlates that strongly. There are too many factors that go into a player's production and that makes correlation fairly difficult to find.
Across the board, the two metrics that impacted the top-24 backs the most were dominator and target share. Here are all the coefficients with respect to both fantasy points and fantasy points per game:
Not much correlation, right? Well, you could say I was initially pretty disappointed. I thought for sure I had come up with some revolutionary combination of analytics to help predict rookie and sophomore running back success in the NFL. How did I come up with the first two years as a benchmark for this model? Well, the average number of seasons that the sampled running backs took to reach RB2 status was 2. Of course, there were some outliers (see: James White), but overall, if a player is going to do it, they're going to do it early.
So after some disappointment, I decided to manipulate the stats a little more. Obviously, dominator and target share had the highest correlation to fantasy output (as we've known for quite some time), but their r-values still didn't signify a positive correlation.
I decided to put together dominator, target share, and burst (the only metrics that had any semblance of statistical significance), giving a composite score of 0.5, 1, 2, or 3. Those numbers indicate how many of those metrics a player was above average at, with 0.5 indicating zero (there's always a non-zero chance that a player will hit). We'll call it the College Composite Score.
The correlation of these three metrics put together as a composite score, plus the metrics that I will talk about soon, with fantasy points per game was 0.50; by far the highest correlation of all this research and the most significant number I reached. And additionally, one of the higher correlations you can realistically achieve when analyzing fantasy football.
So how did I get there?
Photo by Charlie Riedel/Associated Press
Often in the fantasy world, the topic of "talent vs. opportunity" is debated. Many would argue that Clyde Edwards-Helaire was far from the most talented back in the 2020 NFL Draft. But once Damien Williams decided to opt out of this NFL season, perceived opportunity shot Clyde all the way to the top of the running back class if he wasn't already there.
The word perceived is key here; if the Chiefs were to sign Devonta Freeman tomorrow, the perceived opportunity would have to change for Clyde Edwards-Helaire. The Leonard Fournette signing knocked Ke'Shawn Vaughn down. But how was I going to build that into this metric?
The model I decided on is simple and fairly easy to understand.
The best example of this metric would be Austin Ekeler vs. Christian McCaffrey. Headed into the season, the perceived opportunity for Austin Ekeler would have been 0.5; he was splitting time with Melvin Gordon pretty exclusively and there was no reason to believe that would change (other than Gordon's threat of holdout). Christian McCaffrey had no backfield competition and was the clear lead back, so he would have received a Perceived Opportunity Rating of 1.
In an example for this year's rookies, we can analyze Jonathan Taylor vs. DeeJay Dallas. Jonathan Taylor, while in a similar situation to Dallas, has a clear opportunity to be the guy in Indianapolis in 2020. DeeJay Dallas' opportunity is limited by a top-ten running back, a 2019 1,000-yard rusher, and a former first round pick once he returns. He does have the opportunity to get on the field, as he's shown great receiving chops all through training camp. For this, Jonathan Taylor gets a rating of 0.7; DeeJay Dallas gets a 0.5. Is this generous for Dallas? Perhaps. But I'd rather overestimate a role than under, as that will likely provide more accurate results.
The inclusion of Perceived Opportunity is crucial to the accuracy of these results; without them, Jason Huntley would have one of the best chances to finish as a top-24 running back, according to this model. That's not the case.
Photo by Joe Robbins/Getty Images
A fact about the NFL that some tend to forget is that draft capital is crucial to a players success. Are there anomalies? Of course; as there are with anything in the world. Chris Carson. Phillip Lindsay. Danny Woodhead. CJ Anderson. But the average draft capital of top-24 backs in the last five years has been third round. 17 of the 24 (71%) running backs from 2019 had third-round pedigree or higher. 16 of the 24 (67%) running backs from 2018 had third-round pedigree or higher.
The draft matters now more than ever; so I decided to incorporate draft capital into this metric. The method is simple: inverse the round to determine the composite score.
Again, undrafted free agents get a designation of 0.5 because there's always a non-zero chance that a player will hit. Even Kalen Ballage had a non-zero chance before averaging less than 2 yards per carry.
There were ten running backs taken in the first three rounds of the 2020 NFL Draft:
Clyde Edwards-Helaire (1st)
D'Andre Swift (2nd)
Jonathan Taylor (2nd)
Cam Akers (2nd)
JK Dobbins (2nd)
AJ Dillon (2nd)
Antonio Gibson (3rd)
Ke'Shawn Vaughn (3rd)
Zack Moss (3rd)
Darrynton Evans (3rd)
Keep this in mind; it's important.
The Projected Running Back Success Score
Photo by Sporting News/Getty Images
So we've discussed the various factors that I analyzed in order to determine if this model was legitimate. In order to determine the Projected Running Back Success (PRBS) score for the 2020 rookies, I had to mesh them together to determine a composite score.
A perfect PRBS score is 21.0. First-round pedigree, the clear workhorse, and above average in dominator, target share, and burst. To put that into perspective, if extended out to veteran running backs, Christian McCaffrey and Saquon Barkley would have finished with a perfect PRBS in 2018 and 2019 and Marshawn Lynch would have in 2017. That's it.
Here are their fantasy finishes in each of those years:
Christian McCaffrey - RB1
Saquon Barkley - RB10
Saquon Barkley - RB1
Christian McCaffrey - RB2
Marshawn Lynch - RB24
Not a single player outside of the top-24 backs would have finished with a perfect score outside of the top-24 backs.
Here's the next tier of PRBS scores (14+) if applied to running backs in the NFL:
Leonard Fournette - RB7
Todd Gurley - RB14
Melvin Gordon - RB23
Adrian Peterson - RB33
Todd Gurley - RB3
Melvin Gordon - RB8
David Johnson - RB9
Adrian Peterson - RB19
Todd Gurley - RB1
Melvin Gordon - RB5
Leonard Fournette - RB9
Christian McCaffrey - RB10
David Johnson - RB1
Melvin Gordon - RB7
Todd Gurley - RB15
Ryan Mathews - RB31
Adrian Peterson - RB2
Todd Gurley - RB9
Pretty good company to be in, right?
The average PRBS score of top-24 backs over the last five years is 6.78. The average PRBS score of this year's rookie class is 3.06; obviously, not even half of the class will meet the average PRBS of an RB2 or higher.
The 2020 Rookies
It should be noted that JK Dobbins, Zack Moss, and Michael Warren II don't have burst scores, and therefore limited their PRBS accuracy. If I had to bet, JK Dobbins would have met the burst threshold and that would have given him a 4.1 PRBS, finishing just behind D'Andre Swift and AJ Dillon.
As you can see, there's quite a bit to take from this information. First, Clyde Edwards-Helaire is the only player that would be classified into the second tier of players that I mentioned above; though, if Marlon Mack wasn't a thing or Damien Williams was a thing, Jonathan Taylor would be right there with him.
Secondly, only three players finished above the average for a top-24 running back. It is worth noting that there could be a guy like Austin Ekeler - a running back that was perceived to be in a time share whose circumstances changed and he performed - within this draft class. Austin Ekeler's PRBS for 2019 was 0.5. That gives guys like James Robinson, Anthony McFarland, and Jason Huntley the most outside-of-chances to finish as top-24 backs in 2020, even if the odds are extremely stacked against them.
The more likely scenario is that two of the top-three guys on this list makes it into the top-24. I would be shocked if neither Jonathan Taylor nor Cam Akers took over as the lead back for their respective teams at some point in 2020. With such a shortened offseason, it's entirely possible that the shift comes too late into the season to get either of them into the top-24. But I really don't think that happens. Taylor's receiving struggles may limit his upside, and the presence of Nyheim Hines surely poses a threat to an immediate three-down role. But using Nick Chubb's rookie season as a benchmark seems to be the easiest comparison. Nick Chubb finished inside the top-20 in his rookie season. Cam Akers is simply going to standout as the most talented running back in the room, despite the Rams' infatuation with Malcolm Brown. There's a world in which all three of the top rookies in PRBS finish inside the top-24.
There are two names on this list that I feel are low and could definitely contend for a top-24 position if situational factors sway into their favor: DeeJay Dallas and Joshua Kelley. The former has been hyped throughout training camp, making some acrobatic catches and firmly cementing his chance to make an impact in the receiving game. But his college metrics don't favor a top-24 finish, so he finishes lower. Joshua Kelley has already climbed into the RB2 spot on his team's depth chart, jumping Justin Jackson in the process. He's been getting an absurd amount of (justified) hype, and it's well within the realm of possibility that he will assume all of the Melvin Gordon role; a role which saw an RB23 finish in just 12 games.
Remember when I said that draft capital is important and listed the guys who were drafted in the top-three rounds? Cross-reference that with the top guys on the list. It may not be the order they were drafted, but the guys drafted in the first three rounds of the NFL Draft all finished with higher PRBS scores (with the exception of Ke'Shawn Vaughn, because Bruce Arians hates all running backs). When team's invest a high pick in a player, they have more of an opportunity to produce. For example, Damien Harris was taken in the third round of the 2019 NFL Draft. His rookie season, he was unplayable. He's emerged this year as a real threat to Sony Michel's starting job (aside from his hand injury) and was given the opportunity partially due to the high investment in him.
So there you have it. The Projected Running Back Success Model. As I mentioned above, there will be anomalies. I would not at all be shocked if an Eno Benjamin or James Robinson finished as an RB2 in the next two years. I do believe that this model gives the best chance at fantasy success for your roster, though.
The plan is to monitor the accuracy of this model as the season progresses and implement it into my dynasty draft preparation in 2021. Of course, this is a model that will only work post-draft. And even then, it's probably going to be a better tool for redraft or long-term dynasty; Joshua Kelley's Perceived Opportunity immediately following the NFL Draft was not 0.5.
I hope you enjoyed the read and can take some value from this research. It's 100% correct, as all of my takes and research are; until they're not (as was the case with Ronald Jones).
Until next time!