Why Fantasy Football Owners Should Be Paying More Attention to Running Back Regression
It's really easy to look at fantasy football results from one season and use them as the basis of your rankings for the next. This is especially true of top players -- if Jamaal Charles goes nuts one year, the chance that he'll be ranked highly across the industry during the upcoming one is pretty high.
The problem with this line of thinking is that great seasons are great because something spectacular happened. Perhaps a player scored a lot of touchdowns. Maybe he saw an unreasonable amount of volume. Or there's a chance the player was super efficient.
Automatically slotting the best players from one year as the best ones during the next in fantasy football ignores regression. And ignoring regression can get you in a lot of trouble in this addicting game we play.
Take C.J. Spiller's 2012 season. That year, he was a top-five running back in fantasy, rushing for 1,244 yards and adding another 459 through the air. He averaged 6.0 yards per carry, becoming just the sixth player in NFL history to rush 200 or more times and average such a ridiculous number of yards per tote.
Entering 2013, Spiller was understandably a top-three fantasy selection.
He finished as the 27th best running back in fantasy football, playing 15 of 16 games.
You may think this is a dramatic example, but it's really just another instance where regression -- a return to a former or less developed state -- is in play. And as you'll see, if you don't pay attention to regression in fantasy football -- specifically at the running back position -- you could put yourself in an undesirable situation.
Exactly how real is regression at the running back position? That's the high-level question I wanted to find an answer to.
So I dug into the data. More specifically, I looked at numberFire's Net Expected Points (NEP) data. If you're unfamiliar with NEP, it's an advanced metric we use to determine what actually happens on the football field, rather than using simple counting stats. A 10-yard gain on 3rd-and-10 is much more important than a 10-yard gain on 3rd-and-15, right? Well, it should be -- one results in a first down, while the other puts the team in a not-so-great 4th-and-5 scenario. One contributes positively towards NEP, while the other may not.
To get more info on Net Expected Points, check out our glossary.
Over the last 10 years, the Rushing NEP (points added on the ground) per rush average among 100-plus attempt running backs has been -0.03. Why isn't it zero? Because rushing is less efficient than passing -- it's easy to move the chains with a throw, but it's not with a run.
For the purpose of this study, I needed players who were uber efficient. I needed guys who outperformed expectation consistently in order to compare how they performed the following year.
Among these 100-plus attempt runners since 2005, the standard deviation in Rushing NEP per rush was .08. Meaning, one standard deviation away would create a range of -0.11 to 0.05 Rushing NEP per rush.
The lower range isn't all that important here, because that's where the ineffective runners sit. The upper range, however, was used as my cutoff for the efficient rushers. A Rushing NEP per rush of 0.05 was the reference I used to determine if a running back in a given season was efficient or not.
That gave a sample of 85 running back seasons.
Now, let's go back to the original question -- how real is regression at the running back position?
Because we all love fantasy football, I didn't just relate this to Net Expected Points. Instead, I looked at the fantasy football points per touch the group of running backs had during their efficient season, and saw how it changed over the following year. The "fantasy points" scored were standard-scoring points, while the "touches" were determined through a combination of attempts and catches. Fantasy points can come through the air, too, after all.
Just to make sure you're still with me: we're looking at efficient running backs in terms of NEP, and comparing their fantasy point per touch outputs during their efficient season to the next.
Got it? Good. Let's take a look at the results.
I think it's safe to assume a lot of these running backs didn't perform as well on a per touch basis during Year 2 of the study. It makes sense logically, at least -- when a running back is well above average in efficiency, maintaining that efficiency is no easy task.
To reiterate: great seasons are great because something spectacular happened.
What I didn't expect to see was that, of the 78 running backs with "following year data" (seven running backs from 2014 were in the original sample, but they obviously don't have next-year numbers yet), only 11 improved in the fantasy points per touch department. That means roughly 86 percent of running backs performed worse in fantasy football on a per touch basis the year after they were super efficient on the ground.
On average, the sample of running backs saw a drop of 0.15 fantasy points per touch from one season to the next. That means a 200-touch running back -- and remember, that includes receptions, too, so it's not like we're only looking at the highest-volume players here -- is, on average, down 30 standard fantasy points from one season to the next. That may not seem significant, but in standard scoring leagues last year, that was the difference between the 10th- and 20th-ranked running back.
Another interesting point with this data set: there was a moderate correlation between Success Rate and fantasy points per touch lost.
Success Rate measures the percentage of rushes that contribute positively towards a player's NEP. In essence, if a player has good efficiency (which is being defined by a Rushing NEP per rush greater than 0.05), you would hope he also has a good Success Rate. Otherwise, if he doesn't, his efficiency may be skewed by bigger plays, as he's not consistently making positive plays.
Take a look at the bottom five players in the data set in Success Rate, and make note of the number of fantasy points per touch these players lost in the following season.
|Year||Player||Rushing NEP per Rush||Success Rate||FF Points/Touch Y1||FF Points/Touch Y2||Difference|
This is just a small snapshot of the correlation, but among the bottom five Success Rate rushers within the sample, the average drop in fantasy points per touch was -0.22, significantly higher than the average among the entire data set. And as the Success Rate grew, the fantasy points per touch difference from one year to the next was minimized.
Another decent correlation was the degree in efficiency versus the fantasy points per touch difference. What I mean by that is the higher the efficiency -- the higher the Rushing NEP per rush mark -- the bigger the difference seen in fantasy points per touch scored from one year to the next.
Like the Success Rate connection, this makes a ton of sense logically. If a player is incredibly efficient, it's going to be tough to keep that up over a long period of time. As a result, fantasy scoring will suffer.
Here's a list of the top five players in Rushing NEP per rush over the last 10 years (again, minimum 100 carries), as well as their difference in fantasy points per touch from their efficient year to the next.
|Year||Player||Rushing NEP per Rush||Success Rate||FF Points/Touch Y1||FF Points/Touch Y2||Difference|
Before I go on, I want to give a little shoutout to Marion Barber and his 2006 season. If you forgot, the man scored 14 rushing touchdowns that year...on 135 carries. That's a fantasy football owner's dream.
But let this serve as a quick glance to the relationship between Rushing NEP per rush and a dip in fantasy points per touch. The average difference from one year to the next in fantasy points per touch among this group was -0.27, far worse than what we saw from the sample as a whole. In other words, the higher the efficiency, the bigger the drop during the following season.
What This Means for 2015
The more I thought about this study, the more I realized it's sort of like baseball's BABIP (batting average on balls in play) metric.
If you're not familiar, BABIP measures, well, it measures exactly what it says -- the batting average a player accumulates on balls hit in play.
Generally, this BABIP number will hover around .300. Anything significantly higher may mean the player is getting lucky, while anything lower means the batter may have some positive regression upcoming.
But some players are just good. Some can hit line drives consistently, while others are fast and can beat out slow grounders to third. Good players can sustain higher BABIP averages.
The same can probably be said for a study like this. Across the board, names like Jamaal Charles, Maurice Jones-Drew, LeSean McCoy and Marshawn Lynch appeared. And that's because these players are really good at their jobs. They're efficient more years than not because they're incredibly gifted athletes who can maintain a high level of effectiveness on the ground.
So when I show you the seven running backs who hit the highly-efficient mold from 2014, don't assume each one of these players is doomed, especially when you consider who some of them are.
|Year||Player||Rushing NEP per Rush||Success Rate|
Remember, we found two big things with the study above: (1) higher Rushing NEP per rush rates tend to see larger fantasy points per touch drop-offs and (2) lower Success Rates have a mild correlation to larger fantasy point per touch dips.
Because Charles and Lynch have proven to be hyper efficient for multiple years, we can probably ignore them on this list. Meanwhile, Le'Veon Bell and Lamar Miller have both metrics favoring them -- their Rushing NEP per rush is almost as low as it can be to even make the study, and their Success Rates are about average among the sample size. They're probably safe.
That leaves us with Justin Forsett, Jeremy Hill and C.J. Anderson. Interestingly enough, Anderson and Hill were both players who fit this rough mold of a first-round fantasy running back bust, so it's not a delight to see them listed here. However, a point in their favor is that, even if they lose efficiency in 2015, there's a very good chance both running backs will see more volume to make up for it.
They're both risky early-round selections given all this information, but I'm 100 percent most concerned about Forsett. Not only is his Rushing NEP per rush high, but his Success Rate is also really low -- among the 85 running backs in the sample, it's the 18th lowest. And of the running backs with a worse Success Rate than Forsett, the median drop in fantasy points per touch is close to -0.20, which is far worse than the average we saw in the study above.
And guess what? Forsett actually was a member of this study when, in 2009, he rushed for a 0.08 Rushing NEP per rush average and a 47.37% Success Rate. The following year, Forsett saw four more carries (114 versus 118), but his fantasy points per touch rate dropped 0.22 points.
This is standard fantasy football, though. The one thing Forsett has going for him in 2015 is that Marc Trestman is his new offensive coordinator, and Trestman has a history of helping running backs catch passes out of the backfield. If Forsett is able to gobble up targets in the Ravens' passing attack -- which could easily happen given the team's weapons -- he shouldn't be in awful shape in PPR formats.
But as a rusher, given this data, it seems we should be a little hesitant with Forsett this season. And it's all because of regression.