If you've been following the site recently, you may know that I've been doing some regression analysis on fantasy scoring using numberFire's Net Expected Points (NEP) data, and talking about what it could mean for 2016. 

On every play, there's an expected point value an NFL team has for the drive based on yard line, down, and distance. What happens on that play can change the expected point value on said drive. What NEP does is aggregate the values gained or lost on every play into a single, net number. That's Net Expected Points. 

You can read more about NEP in our glossary.

So far I've addressed the quarterback and wide receiver positions. Today, we'll tackle the running backs (pun intended). 

Introducing RB Score

The running backs were a little different to evaluate compared to the other positions. For quarterback and wide receiver, there were particular NEP metrics that directly correlated to fantasy success, Passing NEP and Reception NEP, respectively. Those stats represent a large portion of most of their production. 

However, for running backs, we have to consider both rushing and receiving, since large quantities of fantasy scoring come from both. It would be terribly useless to discuss running back scoring without both of those things, especially since I am looking at PPR point scoring. 

Further complicating things is that Rushing NEP was only loosely correlated to fantasy points per game coming from rushing (r squared of .35) when looking at all running backs with a minimum of 140 carries since 2011. Since we know running the ball is less efficient than passing, this should come as no surprise. Often times, running backs with few rushing attempts were able to see high Rushing NEP totals because a handful of their plays went for big NEP gains. That, however, wouldn't be sustainable over time.

Fortunately, we also track Rushing Successes here at numberFire. A Rushing Success is any rush that yields positive NEP, or an expected points gain greater than zero. 

When I ran a regression on Rushing Successes per game versus rushing fantasy points per game, the correlation coefficient came in much higher, at .7764. 

We already know from the wide receiver piece that Reception NEP correlates strongly with receiving fantasy points per game, so it just became a matter of combining that and the rushing success data.

I now bring to you RB Score, which is Rushing Successes per Game times a coefficient of 9 (best fit) plus Reception NEP.

Here's how RB Score correlated to total fantasy points per game for running backs.

There was an incredibly strong relationship between RB Score and fantasy points per game for running backs in the overall sample (168 players total).

Using the equation at the bottom (with RB Score as "x"), I was able to create an expected fantasy points per game value for each running back based off of their RB Score. Then, all I had to do was subtract that from a running back's actual fantasy points per game to arrive at Running Back Points Per Game over Expectation (RBPPGoE).

Is RBPPGoE Predictive?

As with all of the previous points over expectation metrics, the benefit of RBPPGoE is that it allows you to see the difference between on-field play and fantasy scoring. In theory, runners with a lower expected than actual fantasy points per game did better than their play on the field would suggest, while those with a higher expected than actual fantasy points per game left some fantasy scoring on the field.

But can this tell us anything about the next year's scoring?

Out of the 72 running backs who qualified for study in successive years, about half of them increased in scoring the following season (49 percent). RBPPGoE was only slightly better than random chance at predicting improvement or regression (57 and 59 percent, respectively). 

However, much like the analysis on wide receivers, we saw predictability for regression when we moved further out on the scale. Of the 18 running backs to qualify for study in successive years and have a RBPPGoE greater than 1.5, 14 of them saw a drop in fantasy points per game the following season (78%). 

Again, just like for wide receivers, I had to move out to an extreme so far that I only had one data point in order to improve upon random chance for predicting improvement. 

But since we can see the predictive power for regression, here are four running back regression candidates for 2016.

Mark Ingram, Saints -- RBPPGoE: 4.95

Mark Ingram was by far the highest scoring running back with respect to his expected fantasy points per game. His RB Score was closest to running backs like Charcandrick West (34th in points per game), Eddie Lacy (49th), and Chris Johnson (44th), but he managed to finish as a top six running back. 

Due to his position in a highly potent offense, he may not fall quite as low as those guys did last season, but drafting him as an RB1 would seem unwise. 

Todd Gurley, Rams -- RBPPGoE: 3.13

This one hurts because I think that anyone who watched Todd Gurley can tell you that he's a stud. Based on RB Score, though, he played more like T.J. Yeldon, who was RB18 last season in fantasy points per game, rather than a top-10 running back. This makes sense considering the futility of the Rams offense as a whole (dead last in NEP per play in 2015). If the Rams don't drastically improve on offense in 2016, fantasy owners taking Gurley with a top six overall pick may be disappointed. 

Devonta Freeman, Falcons -- RBPPGoE: 1.90

Freeman graded out awesome last season, with a RB Score over 100. However, achieving a RB Score that high is incredibly difficult, and even more difficult to maintain. 

Out of the 11 running backs to break 100 RB Score, only one of them -- Matt Forte in 2013 and 2014 -- was able to do it again the next season. And of the eight running backs to go over 100 and qualify in successive seasons, only Forte (again) avoided a drop-off in fantasy points per game. 

To frame just how difficult that was for him to do, Forte saw 28 more receptions in 2014 than he did in 2013, but only went up in fantasy scoring by 0.4 fantasy points per game. It simply isn't likely that Freeman sees a similar bump. 

Latavius Murray, Raiders -- RBPPGoE: 1.50

Latavius Murray was an RB2 last season in fantasy points per game, and it appears as though he was lucky to even accomplish that. Especially with Oakland discussing adding a running back to the mix, Murray should probably be avoided in drafts. At least as it stands right now.