A Deeper Look At The Tight End Position Through Clustering
So, after writing my last article about the value of centers in today’s NFL, I wanted to continue with another position. This piece is going to focus on tight ends. I’ll get into the positional value and such for a bit, but I also used K-Means clustering to group tight ends based on a chosen set of variables I’ll cover later in this article.
What’s Their Role and Value?
I don’t have much basis for this, but I don’t feel like tight ends are talked about that much outside of the big name guys like Travis Kelce and George Kittle. Like how often do we hear analysts and fans say to go check out Tyler Higbee or Geoff Swaim. And maybe that’s rightfully so. To the common fan, it may seem that average tight ends aren’t that important. However, I want to bring to your attention some of their responsibility and how they are valued.
The role of the modern tight end has changed a lot from the past. We’re seeing new levels of involvement in the passing game, with some players like Mike Gesicki basically serving as oversized wide receivers. Yet, their main responsibilities are still the same. Block in the run game (and sometimes the passing game) and run routes and catch passes. That sounds a lot simpler than it really is. Assuming a fairly even split between run and pass plays, coaches are asking their tight ends to be an offensive linemen for half of the plays and to be a receiver on the other half. Certainly a lot to ask.
Let’s go back to that seem PFF WAR graphic from the last piece, this time looking at tight ends.
Based on this, tight ends are a decently valuable position, but not one of the super important ones. I’d agree with that as a whole. However, players like Kelce, Kittle, Mark Andrews and such will certainly give you fantastic output every year. One way to show this is through the top-ten average PFF WAR by position:
Once we look at it this way, we see that great tight ends have a lot of value. But at the same time, for every elite tight end, there’s two to three average tight ends to match. It’s a somewhat scarce position, as you usually only see four to five guys above 80 by PFF’s grading each year. To qualify as a great tight end, you’ve got to do one of blocking or receiving at an insanely high level. There’s very few players that can do both. Yet, guys like Nick Boyle, a Ravens tight end, have stuck in this league thanks to their fantastic blocking ability. Elite blocking tight ends can completely turn around run players ito the offense’s favor, serving as a sixth linemen. Excuse the quality, but Boyle (86) takes his defender out of the play here, giving Lamar room to run:
Be a great blocker and you’ll have a place in this league. But, his receiving deficiencies result in him being viewed as an average tight end, which is fair.
If you have a tight end that can do block and receive at an elite level, they will instantly be one of the most valuable players on your team. Other than that, albeit a few good tight ends, most are interchangeable unless they posses a trait they’re amazing at, like the aforementioned Nick Boyle.
One way to see to see the everchanging value of tight ends is through their contracts. For a long time, their contracts were normally under 10 million dollars per year. Then teams begin to utilize them differently and their salaries rose. Now, they’re seven tight ends making over 10 million dollars per year, with four making at least 14 million dollars. When players like Kyle Pitts and Darren Waller become eligible for contracts, I wouldn’t be surprised if they began to push near 16 or 18, even 20 million dollars for an annual value. It’ll still take a little more time, but if we continue to see tight ends with the skillsets of the current elites, their per year value may inch even closer to that of wide receivers.
To put a bow on this section, I’d say that an elite tight end, whether it be in receiving, run-blocking, or both, has immense value to a team but other than that it’s a middle of the pack position that will see players interchanged often. To give myself a better understanding of the different types of tight ends, I clustered them based on different stats from the 2021 season, which we’ll get into now.
Tight End Clustering
This brief introduction to what I did may sound a bit nerdy, but it’s necessary. I really do want to try make this as understandable as possible to everyone, so bear with me. Following this great tutorial from Alex Stern and drawing inspiration from this article from @arjunmenon100, I used K-Means clustering to group together tight ends based on similar traits. K-Means clustering is focused on creating clumps based on the value of predictor or feature variables. To start, I needed to select “features” or stats to use in my cluster analysis and group players off of. Here’s what I used:
- Average Depth of Target
- Yards Per Route Run
- Number Of Targets
- Yards After Catch Per Reception
- Contested Catch Rate
- Drop Rate
- Route Rate (% Of Snaps Running A Route)
- Inline Rate (% Of Snaps Lining Up Inline)
- Slot Rate (% Of Snaps Lining Up In The Slot)
- Wide Rate (% Of Snaps Lining Up Out Wide)
- Block Rate (% Of Snaps Blocking)
- PFF Pass Blocking Grade
- PFF Run Blocking Grade
I thought this group of stats would do the best job of analyzing the receiving, blocking, and utilization aspects that go into grouping tight ends together. We get a look at receiving efficiency, how they obtain that efficiency, alignment rates, and their blocking ability. There’s an argument to be made that some of the stats are redudant, but I went for safety as opposed to leaving something out. From these features, I was able to distinguish six clusters for tight ends.
That’s a lot to take in in one graph. That graph contains the six clusters, and the relative importance of each stat in that cluster. To make it easier to understand, here’s a simpler breakdown of each cluster:
Cluster 1: Elite Receiving Tight Ends
- High aDOT and Yards Per Route Run
- Usually Line Up In the Slot or Out Wide
- Low Block Rate
Cluster 2: Look For Another Option
- Poor Blockers
- Low Receiving Efficiency
- Line Up As A Receiver Often
Cluster 3: Inline Tight Ends
- High Inline Rate
- Decent Blockers
- Low Route Rate
Cluster 4: Versatile Guys, Except For The Slot
- Decent Receivers and Blockers
- Low Slot Rate
- High Contested Catch Rate
Cluster 5: Slot Tight Ends
- High Slot Rate
- Low Drop Rate
- Low YAC Potential
Cluster 6: The Do-It-All Tight Ends
- Great Blockers and Receivers
- Low Wide and Slot Rate
- Great YAC Ability
Hopefully, that helps to understand the main traits these tight ends were grouped off of and get an idea of the different tight end archetypes. Keep in mind, this is only for the 2021 season, and a tight end may be in a different cluster for his full career. I also applied a play filter, ensuring that we don’t get any outliers or players who just didn’t play enough snaps. Nonetheless, this gives us a good look at the ways we can group tight ends based on easy to understand metrics and recent performance.
Next in my analysis, I utilized Principal Component Analysis to visualize these player clusters. This method of analysis allows for datasets to be reduced in dimensionality, making them easier to understand by creating new variables that help to explain the information just the same while miniziming complexity. With that, comes this graph:
This is our first look at which players fall in under which cluster. This also needs the context of which cluster is which, so for that:
Cluster 1 — Top Green | Cluster 2: Bottom Orange | Cluster 3: Right Purple | Cluster 4: Middle Pink | Cluster 5: Middle Green | Cluster 6: Top Yellow
Let’s address some of the wackiness from this graph. Mike Gesicki’s an outlier from his cluster to due to his absurdly low inline rate, as he’s basically a receiver. George Kittle barely fits his cluster due to simply being that much better than everyone else in it. There aren’t many players in cluster 3 or 4 that we see and say “Oh, yeah, they’re a good player”. I’ll let you all draw any more conclusions you see fit.
For another look, perhaps easier, here’s the same thing but in table form:
Now, let’s take a second to analyze each cluster to see if the players in them match up with the traits we defined them with.
Cluster 1: The mainstay receving options at tight end pop up in Travis Kelce and Darren Waller. We also see rookie Kyle Pitts make an appearance, which certainly bodes well for his future in the league. These guys are all uber-efficient for tight ends and don’t line up inline often as well.
Cluster 2: Cluster 2 doesn’t contain many inspiring names. These guys are all average at best. None of them are being paid money, and I’m sure teams would love to upgrade from each. Only a few are starters, such as Dawson Knox and Jared Cook, but who knows for how long.
Cluster 3: This group’s similar to cluster 2 in the sense that they aren’t too inspiring. Austin Hooper makes an appearance in an otherwise lowly group of tight ends, and even Hooper has fell off from his past performance when on the Atlanta Falcons. Most of these guys should be your second tight end, not your first.
Cluster 4: Here we get to see some solid receiving options at tight end. We’ve seen these guys have their moments, like C.J. Uzomah and his two touchdown performance this year, but they still leave a little on the table when it comes to run-blocking.
Cluster 5: It’s nice to see guys like Zach Ertz and Dalton Schultz in this group, given their slot rates. These players may not offer much after the catch, but they are reliable in the passing game. I’m sure Giants will love seeing Evan Engram in a group which is defined by their low drop rate! He’s had his issues in the past, but he’s been alright this year in that department.
Cluster 6: These guys will do a job and then some for your team. Names like Kittle and Rob Gronkowski certainly fit the bill of great blockers and receivers. Perhaps we see David Njoku’s career kickstart with a newteam?
Just based on a quick eye test, it seems the best place to have your tight ends are in clusters 1 and 6. Unfortunately, our good friend Nick Boyle didn’t make it onto the graphics due to an injury he sustained this year.
To wrap this up, tight ends have great value in today’s league, as long as they have the right makeup. We saw that most tight ends we think of as good fall in clusters 1 and 6, where they are either elite receivers or plus blockers and plus receivers.
Clustering also gives a satisfying look at which players are similar to each other, and I’d love to look further into how teams pair their tight ends. Do they look for players from similar clusters? Do they go for different traits? Definitely something I’ll continue to look into. One thing I’d truly love to do would be to apply each player’s PFF WAR to them and see which cluster produces the highest WAR on average. Obviously, due to it being a restricted metric as of now, I can’t, but I think that would go a really long way in our understanding of the position.
For now, though, a large thanks to Alex Stern and Arjun Menon for their articles on clustering and hopefully I was able to leave you with digestable information.
