Welcome back to my NFL Game Script series, where I examine how game script affects running back fantasy production in season-long redraft and best ball formats.
In Part 1 of this series, I examined game script from a bird’s eye view. Let’s recap some of the important findings from Part 1:
- Game condition (leading vs. trailing) has a significant effect on league-wide pass-rates. On average, teams pass about 17% more often when trailing than when they play with a lead.
- Preseason Vegas win totals produce significant correlations with regular season pass and rush totals, which confirms their predictive reliability in projecting a team’s season-long game script.
- I broke our running back sample (n=273) into quartiles based on preseason Vegas win total and reviewed summary statistics for players in each of those quartiles.
- Generally, running backs from Quartile 3 boast the best overall production by most measures.
Quartile 1 is composed of players from teams with a Vegas Win Total of 9.5 or higher. Quartile 2 is 8.5 to 9.0; Quartile 3 is 7.5 to 8.0; and Quartile 4 is 7.0 and lower.
This article will focus on running backs from Quartile 1. But, before I provide recommendations on 2019’s Q1 running backs, I need to take a detour to set up my research methods and results. So, strap in.
Individual Running Back Usage is Fairly Static Across a Player’s Career
In Part 1, I ran correlations between Vegas win totals and regular season rush and pass totals in order to confirm win total’s reliability as a predictive tool. If we intend to project a team’s season-long game script, it’s important that we confirm that our predictive tool (Vegas win totals) is reliable. Similarly, if we intend to project an individual running back’s rushing and receiving usage, we also need to confirm that player usage is predictable.
So, I compiled career rushing and receiving stats for all players in our sample. Then, I calculated each player’s rushing and receiving percentage of opportunities (Opps%) across his career. Let me provide a hypothetical example of what I mean:
Player X has had a seven-year NFL career and has amassed 1,000 rushing attempts and 500 receiving targets over that span. Altogether, his 1,000 rushes and 500 targets sum to 1,500 total opportunities. Of those 1,500 opportunities, 66.7% were rushing attempts and 33.3% were receiving targets. So, his career Opps% (Rush) is 66.7% and his career Opps% (Tar) is 33.3%.
We can perform the same kind of analysis for individual seasons, of course. Let’s say that in 2016, Player Y totaled 200 rush attempts and 50 receiving targets. His Opps% (Rush) for the 2016 season would be 80.0% and his Opps% (Tar) would be 20.0%.
I calculated career Opps% (Rush) and Opps% (Tar) for each player in our sample, and then correlated these career splits with the same Opps% splits from individual seasons. Obviously, for players who only played a single NFL season — or for rookies from 2018 — this correlation would produce a perfect coefficient of r=1.00. So, I excluded these players from my analysis so that they would not distort our results.
As you can see, the correlational coefficients between career Opps% and individual season Opps% are extremely high for each game condition and in total. This means that a player’s rushing or receiving Opps% is reasonably predictable from season to season. Moreover, this further suggests that major changes to a player’s historical rushing or receiving distribution are very rare. If a player has historically been deployed as a satellite receiving back with an Opps% (Tar) of 70.0%, it’s highly unlikely that he’ll suddenly earn 250 carries and only 20 receiving targets the following season.
Alfred Morris’ career stats serve as a clear example of this phenomenon. For each of Morris’ five seasons in our sample, his individual season Opps% nearly perfectly mirrored his career Opps% splits with very little variability from season to season.
This result has certain implications for 2019. For example, Tarik Cohen is likely to maintain his strong receiving usage 1 rather than stealing a large portion of Jordan Howard’s vacated rush attempts. Mike Davis also reports a high Opps% (Tar) throughout his career, which means David Montgomery is likely the best candidate to earn the majority of Chicago’s rush attempts this season.
There are plenty of other examples I could bring up at this point, but instead let’s return focus to our subject of interest. The most important takeaway from these strong correlations is that we can use players’ career Opps% splits to project how a player is likely to be used in any individual season. This is great news. By confirming that player rushing v. receiving usage does not fluctuate wildly from season-to-season, we can isolate game script as the primary determinant of a player’s PPR success based on his “fit” within that projected game script.
Receiving Usage is More Predictable than Rushing Usage
In the previous section, I demonstrated that a player’s style of usage is predictable from season to season. Nonetheless, distribution of opportunities is not the same thing as volume of opportunities. So, my next step was to evaluate how individual season Opps% splits correlate to team target share and team rushing share.
First, I ran correlations for all players in our sample based on each game script condition. As you can see in the graph above, Opps% (Tar) boasts a stronger correlation with team target share than Opps% (Rush) does with team rushing share.
Most strikingly, however, the correlations diverge significantly given a trailing game script. When teams fall behind, a player’s Opps% (Tar) is much more predictive of target share than his Opps% (Rush) is of rushing share. This makes sense, of course. If a team chiefly utilizes a particular running back like Derrick Henry as an early-down back, that doesn’t guarantee him heavy rushing usage in all game conditions. If the Titans fall behind early, Henry’s rushing prowess may not necessarily translate into the lion’s share of rush attempts. Instead, third-down backs like Dion Lewis are more apt to poach rush attempts due to their more versatile overall skillset.
Receiving Importance is Amplified for Elite Workhorse Backs
Next, I ran the same correlations only for players with a team opportunity share of 20% or higher. These players earn either a rush attempt or receiving target on at least 20% of their team’s offensive plays.
As you can see, the correlations for Opps% (Tar) skyrocket while the correlations for Opps% (Rush) plummet. This result suggests that if a player commands a dominant share of his team’s total offensive opportunities — and if that player boasts a high Opps% (Tar) for the given season — then he is all but guaranteed to own a dominant target share for his team. Conversely, the low coefficients for Opps% (Rush) suggest that players who amass an elite share of their team’s opportunities primarily via rush attempts do not necessarily “own” the rushing shares in their respective backfields. These rush-heavy running backs are just as likely to be locked in a committee as they are to operate as true workhorses.
So, if a player possesses an elite receiving skillset, his team is definitely going to utilize him in the passing game. If a player has a predominantly rushing skillset, it’s not a given that he’ll command a dominant team rushing share. This makes rush-heavy running backs particularly susceptible to varying game scripts. It also reinforces that a player’s receiving acumen is more predictive of his future opportunities across all game script conditions.
Last season, James White and Doug Martin earned nearly the exact same percentage of their team’s offensive opportunities.2 One of those players (White) was a redraft lock from week to week. The other (Martin) struggled to maintain any shred of fantasy relevance.
Quartile 1 Running Backs’ Historical Performance
To recap our findings so far:
- Players’ Opps% splits remain fairly consistent throughout their careers. This enables us to reliably predict the style of usage (rushing vs receiving) that a player is likely to earn in his upcoming season.
- Players’ Opps% (Tar) is extremely predictive of projected team target share. By contrast, players’ Opps% (Rush) is not overly predictive of projected team rushing share.
Now, let’s finally turn our attention to Quartile 1 running backs. For all Quartile 1 players in our sample, I correlated players’ team rush share and team target share with end-of-season PPR position rank. This correlational analysis helps us deduce which facet of running backs’ production (rushing or receiving) is more closely related to PPR production for each game condition. Put differently, these results inform us if a player’s rushing or receiving volume is more important to PPR scoring based on whether the player’s team is leading or trailing more often during the season.
I performed these correlations for all Quartile 1 players. Then, I re-ran the correlations for two subsets of our sample population: (1) players with a high share of their team’s total opportunities (defined as 25.0% opportunity share or higher), and (2) players with a moderate share of their team’s total opportunities (defined as 15.0% to 24.9% opportunity share).
As you’re interpreting these results, keep in mind what a correlational coefficient actually is. A correlational coefficient reports the strength of the relationship between any two variables. A coefficient of r=0.40 is considered a moderate relationship, r=0.60 is considered moderate-to-strong and r=0.80 or higher is considered very strong. When the coefficient approaches zero, this signals no apparent relationship between the two variables in question. Moreover, a negative correlation indicates that an increase in one variable (say, Team Rush %) produces a decrease in the target variable (PPR Pos. Rank).
Okay, so what do these correlations mean, exactly? First off, for all players in our sample, team rushing share (aka: Team Rush %) is most predictive of end-of-season PPR position rank no matter the game script condition. This suggests that significant rushing volume is prerequisite to any fantasy success for Quartile 1 running backs. Simple enough.
However, among players with a high opportunity share, this paradigm is somewhat reversed. If a player earns at least 25% of his team’s total offensive opportunities, that player’s target share is far more important to his PPR production than his rushing share is — especially when playing with a lead. Keep in mind that Quartile 1 running backs come from teams boasting elite Vegas win totals. So, we should realistically expect these teams to be playing with a lead more often than not. Therefore, this particular finding discourages us from drafting high-volume rushers without receiving upside who play for elite NFL franchises.
Let’s use another hypothetical example to help illustrate this point. Let us imagine that two players — Player A and Player B — share the same yards per carry (5.0), yards per reception (5.0), total opportunities (250) and team opportunity share (25.0%) for a given season:
Player A rushes 200 times for 1,000 yards and catches 50 passes for 250 yards. His resultant PPR total (excluding touchdowns) is 175.0.
Player B rushes 150 times for 750 yards and catches 100 passes for 500 yards. His resultant PPR total (excluding touchdowns) is 225.0.
Total opportunities is the same, total yardage is the same, but the difference in their PPR totals is entirely due to Player B’s extra 50 receptions.
This logic is not particularly complex: If two players’ opportunity shares are the same, then the PPR format favors players with a higher percentage of receptions. This is especially true for running backs on Quartile 1 teams.
Moving along to players with a moderate share (15.0% to 24.9%) of their team’s opportunities. These kinds of running backs do not command the lion’s share of their team’s total opportunities and are therefore likely mired in a backfield committee. Among players in this range, rushing volume is completely irrelevant. Instead, team target share is the only meaningful determinant of PPR production — whether their team is winning or losing.
2019 Quartile 1 Running Backs
2019 Quartile 1 NFL Franchises: Indianapolis Colts, Kansas City Chiefs, Los Angeles Chargers, Los Angeles Rams, New England Patriots, New Orleans Saints and Philadelphia Eagles.
Below, I have highlighted Quartile 1 players who are likely to excel when playing with a lead and those who are likely to under-perform ADP due to their poor historical Opps% splits. As we examined in the previous section, players with a strong projected rushing and overall workload still require receiving acumen to excel on Quartile 1 teams — rushing alone is not enough. And, for players who do not have a strangle-hold on their team’s opportunity share, rushing is almost entirely irrelevant; instead, PPR success is intimately tied to receiving prowess in all game conditions.
“Can’t Miss” Players
The addition of Latavius Murray likely will not affect Alvin Kamara’s elite PPR production due to his historical team target percentage and versatility. There are zero red flags in his statistical profile that would otherwise signal cause for concern.
Melvin Gordon’s current holdout situation is one to monitor intensely leading up to Week 1, but his historical receiving usage suggests that he will continue to excel if he plays for a contender. If, however, he is traded to the Bucs, Texans — or any other below-average franchise — then his outlook changes significantly.
Todd Gurley’s historical Opps% (Tar) is exceptional for a running back of his magnitude. His on-field usage does not suggest any cause for concern in 2019. However, his overall health and projected volume of touches is another issue entirely. If, whether by injury or by coaching design, Gurley operates with a restricted workload this season, rookie Darrell Henderson has the collegiate profile to excel in Gurley’s stead. Henderson is one of the most prolific college rushers in recent memory, but he also amassed 758 receiving yards and eight receiving touchdowns in his three-year career at Memphis. He possesses the kind of three-down skillset that should translate well playing for a contender like the Rams.
Players to “Buy” at Current ADP
White boasts extreme Opps% (Tar) and target share splits across all game conditions and is a perfect mid-round target. He likely achieved peak opportunity last season 3, but his receiving role in New England’s offense is essentially immutable. As a result, he has an above-average chance to reproduce his 2018 stat-line, which was rich in receiving production.
White exemplifies precisely the kind of “moderate opportunity share” running back we want to target on Quartile 1 teams. He commands a stable role in his offense, he offers upside for a strong opportunity share and his total opportunities are dominated by receiving targets. At ADP RB26, he is currently being under-drafted based on his potential game script.
Austin Ekeler and Alvin Kamara are literally the same player, but in different backfield situations. Just look at each player’s career Opps% splits, heat-mapped based on our entire running back sample:
On top of that, Ekeler and Kamara also share eerily similar workout metrics, both with above average size-adjusted speed and elite explosiveness.
Both players have similar athletic gifts, and through their two-year careers, both have been utilized in almost the exact same way by their respective offenses. Now, Ekeler has an opportunity to seize lead-back duties in L.A. — contingent on Melvin Gordon’s holdout status or a potential trade. Ekeler offers league-winning upside (a phrase I’m hesitant to ever use) and is a no-brainer on draft day.
Players with Uncertain Outlooks
Damien Williams’ career statistics do not paint a black-and-white picture for his potential usage this season. Analysts have speculated that Williams is in line to inherit a large overall workload based on his performance late last season. However, he may still operate in a committee with Carlos Hyde and Darwin Thompson. If Williams earns at least 25% of the Chiefs’ opportunities share, then there is no need for concern. His 49% career Opps% (Rush) when trailing is respectable and may hint at his ability to shoulder the load when Kansas City faces an uphill game script.
However, he doesn’t have any historical track record of handling workhorse duties (career 9.6% team rushing share and 4.8% team target share). Nonetheless, even if Williams finds himself in a committee backfield, his rushing volume would largely be irrelevant for his overall PPR production. His career 33.6% Opps% (Tar) 4 suggests that he would command a very high target share if given sufficient opportunity.
Due to the high likelihood that he achieves a dominant team target share, his fantasy floor likely mirrors a player like James White. However, his poor career rushing workload does not bode well for elite PPR status. His current ADP (RB13) is likely spot-on.
Marlon Mack is a trendy breakout candidate this season — and for good reason. He is projected to be a true workhorse back in Indianapolis, and he proved late last season that he can be a fantasy monster when he’s fully unleashed. However, he has a big problem: Nyheim Hines.
Hines offers the Colts a go-to gadget player who can line up anywhere in the formation and run himself open with his blistering speed. He played both tail back and wide receiver at North Carolina State, and he immediately made good on his receiving skills as a rookie by hauling in 63 receptions on 81 targets. Hines’ presence in the Indianapolis backfield severely caps Mack’s potential receiving usage.
Moreover, Mack doesn’t have a great career Opps% (Tar) to begin with 5. That does not inspire belief that he will suddenly earn more targets this season — especially with Hines lurking around.
As a result, Mack’s fantasy success hinges on his ability to achieve a 25% opportunity share in the Colts’ high-powered offense. Mack managed a 17.4% share last season despite missing multiple games due to injury. It’s certainly possible that he reaches this achievement in 2019, but his historical receiving production is still lackluster at best. Altogether, he best projects as a PPR RB2 without any inside chance at end-of-season RB1 status. At ADP RB17, he may be slightly over-drafted due to his poor receiving upside when playing with a lead.
Players to “Sell” at Current ADP
On paper, Michel seems like a potential fantasy dud. His 5.0% Opps% (Tar) is the second-lowest in our entire sample of 273 players. His 2018 statistics suggest he is completely dependent on rushing volume. It would be one thing if he was the sole option in the Patriots backfield, but instead he shares running back duties with Damien Harris, James White and Rex Burkhead. Given OC Josh McDaniels’ long track record with running backs, Michel is highly unlikely to achieve a 25.0% opportunity share in New England’s offense. That likely places him in the “moderate opportunity share” tier — wherein rushing volume has little-to-no bearing on PPR success.
Drafting a rush-heavy workhorse on an elite team seems like a great idea, but the correlational data suggests Michel’s rushing volume likely still won’t be enough to push him into “breakout” territory. At RB24 off the board as of writing, he may be fairly drafted, but don’t expect an ounce of upside at that ADP.
See: Michel, Sony.
Howard presents a similar worst-case scenario to Michel. Howard is unlikely to see workhorse duty in the crowded Eagles backfield. The most likely scenario is that Howard and rookie Miles Sanders will eventually end up splitting most of the Eagles rush attempts. And, don’t forget that Corey Clement is also still around to snipe some of that Eagles opportunity share as well.
Howard’s backfield competition likely leaves him as a “moderate opportunity share” guy like Michel. And, once again, rushing is mostly irrelevant to PPR success for players in that workload category. His career Opps% splits suggest that he may produce slightly improved Josh Adams-style production, which is not what you want out of your RB3.
Conversely and relatedly, Miles Sanders remains a strong “buy” candidate at ADP RB32. Sanders is much more likely to be serviceable in the Eagles receiving game and also harbors upside to steal all of Howard’s carries by mid-season. If he achieves both tasks, he would fit all positive correlational trends as a potential PPR stud.
What to Expect in Part 3
This installment featured another healthy dose of research methods and exposition, but all that set-up was necessary in order for me to convey my methodology with academic integrity. In future installments, we’ll dive straight into the meat and potatoes for each quartile and spend a little extra time breaking down players rather than a bunch of correlations.
In Part 3, we’ll focus on Quartile 2 running backs, which includes players from the Atlanta Falcons, Baltimore Ravens, Chicago Bears, Cleveland Browns, Dallas Cowboys, Green Bay Packers, Houston Texans, Minnesota Vikings, Pittsburgh Steelers and Seattle Seahawks.