Evaluating Fantasy Football TEs Through Clustering

Introduction

In the dynamic world of fantasy football, understanding player performance and potential is crucial for drafting a winning team. Tight ends are particularly challenging to evaluate due to their dual roles in blocking and receiving, which often results in a wide range of outcomes on the field. Clustering, a statistical method used to group objects that are more similar to each other than to those in other groups, provides a novel approach to analyzing TEs based on various performance metrics. This technique allows us to identify patterns and similarities among TEs. In this article, we’ll explore the different types of TEs in fantasy football and how they impact the game.

Clustering

To cluster, we will be following an unsupervised machine-learning method called k-means clustering. This is where we take all the objects in a group and determine the optimal number of clusters (k) that they can be divided into to best describe the data. For this project, I evaluated TEs by nine different metrics: receiving yards, receiving touchdowns, targets, receiving air yards, air yards share, target share, receiving first downs, and fantasy points per game. These statistics were then converted into per-game metrics.

Using data from the past three seasons, clustering by these metrics generated three clusters of fantasy TEs.

The clusters didn’t surprise much–in fact, there was little variability between the distinguishing factors of each group. In short, we can classify the three clusters as follows:

Cluster 1: Backup TEs

Cluster 2: TE1s

Cluster 3: Waiver wire TEs

The plot below shows the average value for each statistic amongst players in the cluster. The y-axis is normalized between 0 and 1 in order to fit the various metrics on the same graph. When we examine TEs from a fantasy perspective, there are obvious tiers into which they are separated. The best TEs do everything well, the backups are mediocre all around, and the waiver pick-ups are on waivers for a reason.

Labeling the players by their respective clusters, we can now reveal which TEs classify as TE1s–it’s a short list.

Dallas Goedert, Darren Waller, George Kittle, Kyle Pitts, Mark Andrews, Sam LaPorta, TJ Hockenson, and Travis Kelce all fall into the top cluster based on the last three years.

Going solely off the 2023-24 season, this group expands a bit (the cluster number changes due to different data, but the separating metrics remain the same).

Conclusion

All in all, if you’re not getting a top-tier TE, it’s not generally worth it to chase any of the lower tiers–they perform worse in every category, and there are a large number of backup-worthy options. If you want to read more about drafting strategies regarding TEs, check out a previous article where I use AI methods to simulate different strategies!

That’s all for this article, feel free to reach out on Twitter with any questions!