Last year I clustered draft-eligible Wide Receivers (Link to WR Clustering) based on different “play style” metrics from PFF. This year I’ve done the same thing but for RBs! The data is taken from the player’s final season of college. So for the ’23 group – it’s only partial season data, but I’m impatient!
I used K-means clustering using the following variables:
- Breakaway run rate – the percentage of rushes of more than 15 yards
- Yards after contact per attempt
- Missed tackles forced per attempt
- Yards (receiving) per Route Run
- Weight
This resulted in 8 different clusters.
Cluster 1 – Hard to Bring Down Home Run Hitters
- High breakaway run rate and high yards after contact per attempt
Cluster 2 – The Big Boy Plodders
- Heavy, low missed tackles forced per attempt and low yards after contact per attempt
Cluster 3 – Lightweight JAGs
- Small, middling in all
Cluster 4 – Big and Shifty
- The biggest, but also high missed tackles forced per attempt, high yards after contact per attempt, and high-ish yards per route run
Cluster 5 – Elusive Home Run Hitters
- Mid-sized, elusive, good pass catchers, highest breakaway run rate
Cluster 6 – All Rounders with Flair through the Air
- Elite pass catchers, solid weight, and well rounded
Cluster 7 – What Exactly Do You Do Well Again?
- Mid-sized, and otherwise literally the lowest in every category
Cluster 8 – All Rounders with Despair through the Air
- Very similar to cluster 6, but with low yards per route run
To give you an idea of some of the players in each cluster, here’s a table outlining them. We had a tie for cluster 6 – both fairly decent players in my opinion.
I then also used the clustering model to predict which cluster each 2023 Running Back would fall into. See below for results. With not all the 2024 and 2025 running backs seeing significant playing time yet I haven’t presented those names.
Will re-run at the end of the year, but a surprise is Sean Tucker – this may change if he weighs in heavier than the 210 lbs. currently listed. We can see paths to success for Zach Evans and Blake Corum in cluster 1 who has names such as Nick Chubb and D’Andre Swift. Ditto for Bigsby, Charbonnet, Miller, and Chase Brown in cluster 8 which has the Breece Halls and Dalvin Cooks of the world, but also Isaiah Spiller, and Clyde Edwards-Helaire.
I’m not the first to be excited about Jahymr Gibbs, but this really excites me. To be honest, this warrants a look at all the NFL Cluster 6 running backs, as it’s really a very successful list.
All in all, a fun little exercise that is definitely more of a “for fun” analytics-style project as opposed to searching for something with a strong signal for future success. Success comes in a large range of shapes and sizes so it wouldn’t surprise me to see a running back from any cluster succeed (although let’s be honest, except cluster 3 or 7).
Feel free to reach out to me on Twitter or the Campus2Canton discord if you’d like to know what cluster any 2015-2025 running back falls into, as obviously I haven’t been able to post all of them here.