Modeling success for running back draft prospects

Recently, I wrote a couple posts on a pre-combine NFL success model for wide receiver prospects, detailing the methodology and results for the 2016 class. Some consensus top prospects – like Laquon Treadwell and Michael Thomas — fared poorly in the analysis, while receivers further down early draft boards found themselves near the top — such as Leonte Carroo and Hunter Sharp. The model wasn’t meant to give you definitive rankings, but instead inform you of how similar receivers from the 2016 class are to prior prospects that found early success in the NFL.

This post is going to use a similar methodology to develop a NFL success model for running back prospects. Again, we’re going to use a logistic regression model that aims to answer a binary questions: Will a particular prospect find early NFL success?

You can define success many ways, but I’m choosing to use a top 12 fantasy point season (PPR) for running backs. The model’s dependent variable of early NFL success is whether or not a player had a top-12 season within his first three years in the NFL.

Since this is pre-NFL combine model, we only used age and production data to train and test the model. The model used 330 running back prospects that entered in the NFL from 2000-2013, splitting the data roughly 2-to-1 into training and testing sets.

After plugging plugging more than a dozen different production statistics into the model and slowly taking away, one-by-one the least statistically significant, we were left with three that provide the most explanatory and predictive power:

1. Draft age

2. Final season rushing touchdowns

3. Final season receiving yards

As you’d expect, the model favors younger prospects, and those with lots of rushing touchdowns and receiving yards. This might seems like a an overly simplistic model, but simplicity should be the goal of every model, as long as you’re not severely compromising accuracy. The model’s accuracy rate for correctly predicted early NFL success or failure on the test set was nearly 85 percent, roughly equal to using NFL draft position as the independent variable.

Here are the top 10 prediction scores for the entire 2000-2013 data set.

The “Predict” column gives the model score (between 1 and 0) for each prospect indicating the likelihood of a top-12 PPR season in a prospects first three years, and the “Top 12” column indicates whether or not the receiver actually had a top-12 PPR year in his first three seasons. Draft position is listed in the table only for reference; it was not part of the model.  

The 50 percent hit-rate for the top-10 backs isn’t overwhelming, but impressive when you consider that draft position wasn’t part of the analysis. The draft positions for our top-10 backs range from 2 to 230, with a median of 54 (roughly the late second round). Opportunity is one of the keys to early NFL success for running backs, and opportunity is certainly highly correlated with how much draft capital was spent on a prospect. So, we shouldn't expect our draft-agnostic prediction scores to align perfectly with early NFL success.

Brian Calhoun is an interesting name to be tied of the highest prediction score. The undersized collegiate workhorse was drafted by the Lions in the third round of the 2006 NFL draft, tore his ACL during his rookie campaign, and then wasn’t heard from again. Calhoun’s disappointing NFL career adds to the failure narrative surrounding Wisconsin running back prospects, including Ron Dayne, Montee Ball and, perhaps, Melvin Gordon.

There aren’t many high-profile busts near the top of the prediction scores – even Trent Richardson had a productive rookie season. The model correctly identified a few running backs that ended up having much more than a single top-12 season, like Steven Jackson, Ray Rice, and LeSean McCoy. None of these three backs were particularly high drafts picks (23 backs in the 2000-2013 pool had even a better draft position than Jackson), but all were young, extremely productive in college, and brought passing game ability to the NFL.

Kevin Smith had had one of the greatest statistical seasons in college football history, accumulating 2,567 rushing yards (third most in history) 29 rushing touchdowns and another 242 yards receiving. Smaller-school status likely led NFL scouts to grade Smith's production on a curve, as he wasn’t selected until the third round of the NFL draft. But, Smith was instantly productive at the next level, with nearly 1,200 yards for scrimmage and eight touchdowns his rookie season (finished as the RB16 in PPR formats, just outside the top-12).

Kerwynn Williams and Ronnie Hillman are currently in the NFL, but only Hillman has sustained success for even a partial season. For Williams, who was productive in limited touches the last couple years (averaged 4.9 yards per carry on 80 attempts), his seventh-round draft position may have been a bigger obstacle to NFL productivity than a lack of talent.

Now that we have our pre-combine, pre-draft model for predicting success in place, in the next article I will apply the model to 2016’s running back draft class.

In the meantime, below is a sortable database that you can explore and see how all the running backs who entered the NFL from 2000-2013 fared in the model.

Kevin Cole is a Lead Writer for PFF Fantasy. You can follow him on Twitter at @Cole_Kev

[table id=1168 /]