Pre-NFL Draft 2023 Running Back Prospect Model

Using college stats, combine metrics and estimated draft positions to project the 2023 running back class into the NFL

In this analysis of 2023 NFL Draft prospects at the running back position, I’m going to detail the quantitative process behind the results, including the importance of different features and how estimated draft position (via GrindingTheMocks), early declare status, athletic traits and production markers all blend into a single prediction. In this case, the prediction will be focused on fantasy football results, specifically the number of top-6, -12, -24, and -36 weekly fantasy finishes over a prospects first three NFL seasons. We could extend the window further to cover a player’s entire career, but the point in the real NFL draft and fantasy rookie drafts is to see a return on investment earlier rather than later.

While my research shows that NFL Scouting Combine results have a significantly greater impact on draft position than actual NFL value and production, you can’t afford to fade the crowd coalescing around their favored prospects. Despite tremendous advancements in our ability to collect data on college prospects, the most meaningful single variable in determining NFL success is draft position. 

This analysis will cover all running backs from the 2023 class currently with an estimated draft position in the top-250, though many of those estimate are unstable beyond the top handful of prospects.

METHODOLOGY

This analysis uses what is known as an ensemble model, or a blend of the results from different models which are then blended together to form a single, and hopefully more precise, final prediction. One of the models is tree-based, the other linear.

The features for each running back in the models are as follows, ordered by relative importance and statistical significance. The stats are from the prospects’ best statistical season in which they played at least five games and logged at least 25 carries. All of the numbers are calculated only using games in which the prospect played, and all numbers are from each prospect’s best statistical college season:

• Draft position

• Share of team receptions

• Share of total (rushing & receiving) yards

• Share of total touchdowns

• Share of rushing attempts

• Weight

• 40-yard dash time (Pro Day discounted 0.03 seconds)

Draft position is the most important feature in predicting fantasy success for running backs, as is encompasses the total evaluation of each player and can dictate NFL opportunity. Looking specifically at the calculated importances in the tree-based model, draft position is more than double any other feature (more significant than for my wide receiver prospect model), with team reception share the second most significant feature. Size and speed get a ton of hype during the combine cycle, but aren’t particularly important for future NFL success, outside of boosting draft positions.

Below I detail the percentile scores for the model (percentile for top fantasy week projections among the last several years of prospects), and the exact stats for each prospect. There is really one near elite tier of prospects this class, with another few names showing promise, and then most of the rest somewhat interchangeable. The randomness of the draft and who falls into opportunity in the NFL are most important for the latter group.

2023 RUNNING BACK CLASS

No surprise here for who’s on top of the rankings. Bijan Robinson is the latest “generational player”¹ who has been consistently ranked in the top-5 prospects across all positions among evaluators with little regard for positional value. Either the NFL or mock drafters are getting smarter though, with Robinson’s estimated draft position closer to 20th overall than the top-10.

My model is trying to predict NFL production, not value over replacement or versus a theoretical opportunity cost. While I wouldn’t take Robinson in the first round, someone will, and his numbers are outstanding in terms of forecasting strong NFL usage. Robinson’s model projection is around the 98th percentile for all prospects since 2012, trailing only Saquon Barkley (100th), Chrisitian McCaffrey, Ezekiel Elliott and Leonard Fournette (all 99th). All the running backs with better scores had actual draft positions significantly better than the 19th slot estimated in this model, reflecting how good Robinson is outside of draft assumptions.

The one question mark on Robinson’s profile is relatively weaker receiving volume, but that appears to be more about scheme than ability. Texas throws around 50 passes to all running backs per season, not leaving much for a back with good hands.

I was a bit surprised that Jahmyr Gibbs is also in the 90th percentile, but his estimated draft position is in the early second round, and his receiving production has been outstanding. Gibbs gathered nearly a quarter of all of Alabama’s receptions, and he mitigated size concerns by running a blistering 4.36 forty at the combine. I don’t think Gibbs will have touchdown upside in the NFL, but he can be an very valuable player in PPR and half-PPR formats, and with “real” NFL value converting high leverage third downs as a receiver.

Gibbs’ projection is probably more fragile than Robinson’s. I think the 20th pick is close to Robinson’s floor, whereas Gibbs sliding another 10-20 spots wouldn’t surprise me. Gibbs is probably locked in at RB2 in the draft and my model, but potentially at low at the 80-85th percentile.