Using Clustering in Evaluating WRs: The Potential Future of Scouting

The wide receiver position is notorious for its vast amount of flavors and varying usage across different offenses in the league. These two facts make evaluating wide receivers much tougher since you’re comparing apples to oranges (a 6’5” 220 X receiver to a 5”10” 160 slot receiver) in a market where everyone has their own favorite fruit (NFL teams unique preferences from the WR position). So, how does one go about separating and sorting players to get a clearer evaluation? Enter clustering.

I won’t bore you by explaining what K-Means clustering is. Just know it’s math stuff that groups data into K numbers of clusters. The 5 variables used to determine a prospect's collegiate “role” were Average Depth of Target (ADOT), Height, Weight, Percentage of Snaps in the Slot, Percentage of Snaps out Wide. The math did its thing turning the data into 9 unique clusters, or groups, of shared characteristics within each cluster that we’ll see below.

I want to be clear that the clusters themselves have limited inherent use since they only show similar players in those 5 metrics. The goal is to figure out what type of wide receiver a prospect is, then use film and analytics to look at the players within the type to figure out what matters. The clustering is just sorting out the different metaphorical fruits, not telling you which fruits are the best. I’ll be doing that throughout the draft process in a series of articles.

The data used featured 153 wide receivers from 2017 to 2022 including the fourteen 2023 prospects mentioned in this article.

Jaxon Smith-Njigba and Kayshon Boutte

Cluster 7 Description: Short depth of target, predominantly plays in the slot, about average size

Cluster 7 Averages:

ADOT: 8.6 Slot%: 87.6% Wide%: 10% Height: 71.5 Weight: 199.7

Cluster 7 Players:

Curtis Samuel, K.J. Hill, Christian Kirk, Amari Rodgers, Parris Campbell, Kadarius Toney, Devin Duvernay, Khalil Shakir, Freddie Swain, Kyle Philips, Justin Jefferson, Velus Jones Jr.

Xavier Hutchinson

Cluster 6 Description: Short-intermediate depth of target, predominantly plays out wide, above average size

Cluster 6 Averages:

ADOT: 10.2 Slot%: 27% Wide%: 72.2% Height: 73.2 Weight: 211

Cluster 6 Players:

ArDarius Stewart, N’Keal Harry, Corey Davis, Brandon Aiyuk, JuJu Smith-Schuster, Bryan Edwards, Zay Jones, Donovan Peoples-Jones, D.J. Moore, K.J. Osborn, Trey Quinn, Laviska Shenault Jr., A.J. Brown, Van Jefferson, Deebo Samuel, David Bell, Juwann Winfree, Drake London, KeeSean Johnson.

Jordan Addison, Zay Flowers, and Jayden Reed

Cluster 2 Description: Intermediate depth of target, predominantly plays out wide, smaller

Cluster 2 Averages:

ADOT: 11.2 Slot%: 27.2% Wide%: 72.1% Height: 70.7 Weight: 184.5

Cluster 2 Players:

Dede Westbrook, Henry Ruggs III, John Ross, James Proche, Anthony Miller, Amon-Ra St. Brown, Ray-Ray McCloud, Anthony Schwartz, Andy Isabella, DeVonta Smith, Diontae Johnson, D’Wayne Eskridge, Marquise Brown, Garrett Wilson, Darnell Mooney, Jahan Dotson, Skyy Moore

Josh Downs and Nathaniel Dell

Cluster 5 Description: Intermediate depth of target, predominantly plays in the slot, smallest

Cluster 5 Averages:

ADOT: 11.3 Slot%: 82.5% Wide%: 17.2% Height: 69.4 Weight: 175.5

Cluster 5 Players:

Braxton Berrios, Elijah Moore, Hunter Renfrow, Jaylen Waddle, Mecole Hardman, Tutu Atwell, Scott Miller, Wan’Dale Robinson, KJ Hamler

Jaylin Hyatt

Cluster 9 Description: Intermediate depth of target, predominantly plays in the slot, above average size

Cluster 9 Averages:

ADOT: 11.4 Slot%: 75.6% Wide%: 19.6% Height: 73.3 Weight: 199.7

Cluster 9 Players:

Cooper Kupp, Lynn Bowden Jr, DaeSean Hamilton, Tyler Johnson, Russell Gage, Rashod Bateman, Jauan Jennings, Terrace Marshall Jr, Jerry Jeudy, Samori Toure, Treylon Burks

Tyler Scott and Marvim Mims

Cluster 1 Description: Intermediate-Deep depth of target, predominantly plays out wide, about average size

Cluster 1 Averages:

ADOT: 13.5 Slot%: 16.1% Wide%: 83.3% Height: 72.5 Weight: 194.8

Cluster 1 Players:

Chad Hansen, Riley Ridley, Chad Williams, CeeDee Lamb, Isaiah Ford, Jalen Reagor, Taywan Taylor, Quez Watkins, Antonio Callaway, Quintez Cephus, Calvin Ridley, Dax Milne, Cedrick Wilson Jr, Ihmir Smith-Marsette, Dante Pettis, Ja’Marr Chase, Deon Cain, Chris Olave, Michael Gallup, Jalen Nailor, Olabisi Johnson, Jameson Williams, Romeo Doubs

Quentin Johnston and Cedric Tillman

Cluster 3 Description: Intermediate-Deep depth of target, predominantly plays out wide, biggest

Cluster 3 Averages:

ADOT: 14.2 Slot%: 10.4% Wide%: 89.3% Height: 75.5 Weight: 219

Cluster 3 Players:

Amara Darboh, Kelvin Harmon, Kenny Golladay, Miles Boykin, Mack Hollins, Travis Fulgham, Mike Williams, Chase Claypool, Noah Brown, Collin Johnson, Auden Tate, Isaiah Hodgins, Courtland Sutton, Michael Pittman Jr, Equanimeous St. Brown, Tee Higgins, Javon Wims, Ben Skowronek, Marcell Ateman, Nico Collins, D.K. Metcalf, Christian Watson, JJ Arcega-Whiteside

AT Perry

Cluster 8 Description: Deep depth of target, predominantly plays out wide, bigger

Cluster 8 Averages:

ADOT: 16.5 Slot%: 11.2% Wide%: 88.6% Height: 73.9 Weight: 201.5

Cluster 8 Players:

Chris Godwin, Terry McLaurin, Josh Malone, Denzel Mims, Josh Reynolds, Gabriel Davis, Damion Ratley, John Hightower, DJ Chark, Dyami Brown, James Washington, Josh Palmer, Marquez Valdes-Scantling, Alec Pierce, Tre’Quan Smith, George Pickens, Darius Slayton, Tyquan Thornton

Conclusion

While the results are not perfect, I believe clustering WRs could be a crucial part of the evaluation process of the position. I am not a numbers or analytics guy by any measures; I would love if someone smarter than me in this field was to improve and build upon my initial idea. I’m happy with the results and look forward to testing out this procedure throughout the upcoming draft cycle. I want to reiterate that these clusters are comparing players of similar roles—as defined by the 5 variables—not comparing players of similar talent, grade, etc.

When it comes to draft content, I think the process of arriving at a conclusion is often more interesting than the conclusion itself. Hopefully, you enjoyed this article and the look into how my evaluation process works and evolves. I encourage you to leave any ideas, suggestions, comments, or feedback as a comment or DM me on Twitter @ThompsonNFL.