The Model
One of the important things to understand about regression analysis is that the variables used to make predictions about the dependent variable interact with each other, and can either help or hurt the resulting model. Multicollinearity, for example, is a problem that can sap the prediction power of a model in some cases. Ultimately, it is up to the person making the model to make decisions on whether or not these cases can be beneficial to the model or not.
But what does this have to do with football? And specifically, what does this have to do with running backs? Well, in my running back model, the vertical leap is actually a negative number. As in, the higher a running back is able to jump, the lower estimated performance in the NFL. To understand why this is happening, let’s look first at how the vertical jump projects prospects on it’s own, and how it interacts with draft position:
This is a linear regression model only using vertical leap as a predictor of NFL performance. There is almost no correlation with an r^2 value of 0.0298 and a t stat of -1.653 .
Similarly, this is a linear regression model only using draft position to predict NFL success. As you might expect, draft position is very (relative to football standards) significant predictor of success, with a r^2 value of 0.3145 and t stat of -6.39.
Now, this is a model with both draft position and the vert. This model’s r^2 value is the highest with 0.3783. Additionally, the t stat of both variables goes up to -3.005 and -7.023 respectively. When added in with draft position, vertical leap went from insignificant to very significant.
But what does this mean? It means, at least according to the data, that the NFL is currently overdrafting based on their explosiveness. Guys like David Wilson, Kenny Irons, and Mikel Leshoure were drafted at least in part due to their superior jumping ability, while guys like Jamaal Charles, Lesean McCoy, and Ray Rice may have been ignored due to their weak jumps. It’s certainly not the end-all be-all, but it should be able to help us avoid some of these busts and find some of these sleepers.
With that in mind, here is the complete running back model:
MS.Yards is market share of team yards (rushing yards gained divided by total team RB rushing yards) and break out age is the age a prospect first records 100 carries. All the other variables are self explanatory, with draft position, height, 40 time, and vertical jump rounding out the model. The r^2 of the model using the training data is 0.4742. When applied to the testing data, the model gives us a r^2 of 0.3834, less than the training data but still a significant correlation.
2017 Class
*Note: These projections are using estimates for combine drills. My final rankings will be different from these*
Dalvin Cook, Florida State
Generally regarded as a top prospect, the model also loves Dalvin Cook. Breaking out at a young age (19) and posting 71.7% market share of yards, Cook is what you look for in a prototypical first round running back prospect.
2. Christian McCaffrey, Stanford
While it may appear as though McCaffrey took a step down from his Heisman-contending year in 2015, the advanced metrics were just as good. In 2015 he put up a 72% market share of yards and followed that up with 70.3% in 2016. McCaffery is a great prospect in the late 1st round/early 2nd round area.
3. Leonard Fournette, LSU
This may initially seem like a shot at Fournette, but it really says more about how great this class of running backs are. If you placed Cook, McCaffrey, and Fournette into last year’s draft, they all would have been sandwiched in between Ezekiel Elliott and Derrick Henry as the top prospects in the draft. These top 3 guys will separate themselves at the combine, a place where Fournette is expected to dominate the 40.
4. D’onta Foreman, Texas
Foreman is the big sleeper of the class. Currently projected as a late 4th/early 5th round pick by CBS Sports (and I expect him to go higher), Foreman has the potential to be a day 1 bellcow for a team. He clocked in a whopping 84% of his team’s rushing yardage, the highest in the entire class. He is held back by the fact that he was a late bloomer, and he may not run as fast as the the top 3, but should be well worth his projected price tag.
5. Joe Mixon, Oklahoma
Another big back like Fournette and Foreman, Mixon is projected to run extremely fast for a guy his size. Mixon also has potential upside in the fact that he had to share the backfield with another highly regarded prospect in Samaje Perine, possibly suppressing his market share of rushing yards (46.8%). Package all of this in with his character concerns, and Mixon is the definition of a boom or bust prospect.
