Who was the best Defensive (LB, ED, DI) player in the NFL in the 2023 Season?
By: Spence Purnell, Co-Founder
Date: September 8th, 2024
By: Spence Purnell, Co-Founder
Date: September 8th, 2024
TL;DR - We analyze 13 defensive metrics, including Pass Rush Win Rate and Double Team Rate to construct an index for the most statistically productive player of the 2023 season. T.J. Watt was the #1 overall producer according to our index
Today, we will use the proprietary SSAT method to explore a (way too late) debate about the Defensive Player of the Year (DPOY) in the NFL in 2023. We analyze 13 performance metrics, including Pass Rush Win Rate and Double Team Rate, to create a "Final Score," which shows the player's overall output for the season. We limited our analysis to the Edge, Linebacker, and Defensive Interior because these positions collect relatively similar stats, but also calculated scores by position so that averages in other positions wouldn't affect the scoring.
It's explained in more detail on our "How the Index Works" page, but the SSAT method constructs an index using a statistical method called standardization. In a standard score (z-score), numbers are compared to their respective averages and then assigned a positive score ranging up to 1.0 if they deviate above average, a negative score up to -1.0 if they deviate below average, and a 0.0 if they are precisely average. This creates a "standardized" unit that allows for index creation across differently scaled metrics, like tackles, which range from 0 to 111, and INTs, which range from 0 to 4. Our patented weighted system allows for more balanced comparison across metrics with such different ranges. Here's the equation for calculating standard scores for each "metric", such as INTs, Tackles, etc.
Weighted Standard Score = ((Player Stat for Metric - Metric Average) / Standard Deviation of Metric) * Weights
We ran the standard score equation for every player for the following stats and then summed them into categories:
Pass Defense Score
Interceptions
Interception Touchdowns
Pass Deflections
Playmaking Score
Forced Fumbles
Fumble Recoveries
Fumble Touchdowns
Run Defense/Tackling Score
Solo Tackles
Tackle Assists
Tackles for Loss (TFL)
Pass Rushing Score
Pressures (Hurries + QB Hits)
Pass Rush Win Rate
Sacks
The Debate
As mentioned, we try to use the index to settle a debate about the DPOY, which was influenced by advanced metrics such as PFF's "Pass Rush Win Rate" (PRWR) and the NFL's "Double Team Rate" (DTR). PFF argued that PRWR, whether it resulted in a statistical outcome (tackle, sack, etc.), still influenced the game because it may draw the quarterback's or other free blockers' attention. We included PRWR as a metric alongside sacks and pressures to gauge how often and impactfully a player entered the backfield in their pass rush score.
In an interview with Micah Parsons, he argues that a player's DTR should be considered alongside their stats, as double teams limit their ability to impact the game statistically. Parsons experienced a league-high 35% DTR, noting that T.J. Watt only received a 14% DTR, and DPOY Myles Garret had a 29% DTR. To accommodate these concerns, we modified the overall Final Score (which includes all stats measured) by multiplying it by the DTR and then adding that number back to the original score, essentially giving the player a % boost to their final score equivalent to their DTR.
Final Score = Weighted Standard Score1 + Weighted Standard Score2 + Weighted Standard Score3... etc...
Final Score = (Final Score * DTR) + Final Score
We then sum the weighted scores into the four category scores and then sum those into an overall Final Score. We also calculated the weighted scores by position so that the averages were not affected by other positions that don't score well in some categories (Edges have a much higher sack average than Linebackers, for example). This allows us to calculate performance scores compared across all three positions and within specific position groups. We then re-scaled these scores from 60-100 so that:
60s = below-average performance
70s = average performance to above average
80s = high performance to elite
90s = exceptional performers
100 = top performer
Let's explore the data of a specific player to help understand. B.J. Hill from the Bengals provides a nice case study on how the index works. As a DI, Hill's 2 INTs and 5 Pass Deflections are closer to average when compared overall, netting him a 79.21 Pass Defense Score. With other relatively strong metrics, Hill comes out 72nd overall with a Final Score of 76.25, well above average. The cards are colored so that red is below average, yellow is average, blue is above average, and green is elite.
However, since DI's only averaged 0.05 INTs (compared to 0.21 overall) and 1.37 pass deflections to 1.90 overall, Hill's performance in these stats within his position group was highly remarkable, showing great coordination and athleticism on both interceptions. He earned the #1 ranking pass defense score for interiors. As a result, his overall rank improved to #8 within his position group, given that the rest of his stats were solid. Because the average INTs and defections for DIs are lower than other positions, Hill's two interceptions and 5 deflections stand out even more, resulting in a high positional score, while the rest of his stats hovered around average to below average for DIs and the league.
Data Notes:
Since we could not secure full access to ESPN's DTR data, we imputed values using the publicly available data to create a prediction model.
We created several filtering and imputations for PFF's PRWR, since some players with very few pass rush snaps had high PRWR. We imputed values for those low snap count players instead of using the PFF data to create a more accurate picture of PRWR for high usage players
Our analysis excludes primarily pass-defense positions like Safety and Cornerbacks (that analysis is forthcoming in the next ARS post) and, therefore, only includes DE, LB, and DT - positions primarily focused on play around the line of scrimmage and nearby. We then excluded a couple of metrics aimed at grading DB performance, like “Passer Rating when Targeted” and "INT Return Yds." However, we did include “Interceptions,” “INT Touchdowns,” and “Pass Deflections,” as these were more common to record, as LBs play some coverage and DE/DTs deflect passes from the line. We filter out players who played in less than 12 games, often considered the benchmark for a fully active season. All of these decisions affect the metric averages and, therefore, the calculation of the rankings.
Still in Beta, please forgive any slow or odd answers