K-Means Clustering helps break down recent offensive trends: Using metrics like EPA per play and PFF grades, offenses are grouped into tiers to better understand team strengths and weaknesses from the past three seasons.
Scheme diversity doesn't always mean better results: While teams like the Ravens thrive with varied looks, others like the 49ers and Chiefs dominate despite being more predictable in their schematic approach.
Subscribe to PFF+: Get access to player grades, PFF Premium Stats, fantasy football rankings, all of the PFF fantasy draft research tools and more!
Estimated Reading Time: 8 minutes
As we move deeper into the NFL offseason, it's a good time to look back at how offenses have performed in recent years and what each team tends to favor schematically. And PFF has a treasure trove of data and advanced metrics that offer a clear lens into the offensive trends shaping the league.
To make sense of it all, we’re turning to one of our go-to tools: K-Means Clustering. This algorithm takes a wide range of inputs — such as expected points added (EPA) per play, PFF grades and more — and sorts teams into “clusters” based on shared traits. Each cluster groups similar offenses, allowing us to draw key insights and identify larger trends across the NFL. It’s one of the most effective ways to transform a complex data set into something digestible and actionable.
Once we feed our selected variables into the clustering algorithm, we’re left with a set of distinct offensive clusters (or tiers).
To establish a hierarchy among them, we’ve ordered the groups based on their average expected points added (EPA) per play. At the top of the list are the Buffalo Bills and Baltimore Ravens, who, while falling short of a Super Bowl berth, have consistently fielded the league’s most effective offenses over the past three seasons.
As we move down the tiers, flaws become more pronounced, whether due to inefficiency, inconsistency or personnel limitations, keeping those teams from joining the elite offensive groups.
Examining the cluster or tier averages across select offensive metrics, the gap between the top two groups is relatively narrow. Teams like the 49ers, Lions and Eagles have benefited from strong offensive line play over this stretch, but they have been slightly less efficient on the ground than the Ravens and Bills.
Since 2022, Tier 3 has included some of the NFL’s highest-graded quarterbacks, with Joe Burrow earning a league-best 93.5 passing grade, Patrick Mahomes ranking fourth at 90.9, and Justin Herbert placing fifth at 89.8. Baker Mayfield has also been productive during that span, throwing 87 touchdown passes, the sixth-most across regular and postseason games.
Still, the consistent quarterback play in this tier has often been held back by underperforming ground games.
When plotting each team’s pass and rush expected points added (EPA) per play, the teams in this tier all fall above the trend line, a sign that their run games have not complemented their passing efficiency.
The Buccaneers may have addressed that imbalance by finding a solution in Bucky Irving, improving from a league-worst 3.5 yards per carry in 2023 to the third-best mark at 5.2 in 2024.
Using all of the variables included in the clustering algorithm, we can generate a final strengths and weaknesses table for each tier. Strengths are defined as metrics where the tier ranks in the top two across all clusters, while weaknesses are those where it ranks in the bottom two. While these summaries won’t apply perfectly to every team — there are always outliers — they serve as accurate representations of the clusters as a whole.
Offensive Scheme Analysis
To wrap up our offensive roadmap, we can examine scheme trends using a method called principal component analysis. This technique reduces a broad set of variables — including play-action rate, RPO rate, usage of shifts and motion and run concept frequencies — into a smaller set of “principal components.”
By plotting the first two components, which account for the most variance in the dataset, we can visualize how similarly (or differently) teams operate in terms of offensive scheme.
The arrows indicate the directional influence of specific variables on team placement. For instance, shift/motion rate and outside zone rate push teams further to the right on the plot. Teams that fall on the far right — such as the Dolphins, Falcons and Vikings — typically come from the Shanahan/McVay coaching tree, along with the 49ers and Rams themselves. As expected, this offensive lineage is known for emphasizing wide zone runs and extensive use of presnap shifts and motion.
Another lens through which we can evaluate offensive schemes is Shannon Entropy, which measures how predictable an offense is based on schematic tendencies. For example, a team that operates exclusively out of shotgun formations would be considered more predictable than one that blends under center, pistol and shotgun looks. To make things even more interesting, we can plot offensive predictability against efficiency to explore whether more diverse schemes lead to better results.
As we can see, there’s little correlation between offensive scheme predictability and overall efficiency over the past three seasons. That leaves us with two terrifying propositions: facing the highly efficient and schematically diverse Ravens, or trying to stop teams like the 49ers, Chiefs and Lions — offenses that are less diverse but equally dominant.
With this, we’ve built a comprehensive roadmap of NFL offenses in recent years. As coaching staffs evolve and coordinators move into new roles — such as Ben Johnson taking over in Chicago — this framework offers a valuable tool for forecasting what changes we might see heading into the 2025 season.