An analytics primer for NY Jets fans: Behind the numbers of football

The age of analytics in the NFL is here to stay. But what is the mystique behind the metrics?

This article is a collaborative effort between Jets X-Factor’s Rivka Boord and Michael Nania

What is your opinion about NFL analytics?

Ask two football fans, and you’ll get three opinions.

Some believe it has revolutionized the game. Others maintain it takes the heart out of the sport they love.

However, like it or not, analytics won’t go away. The trend made famous by Moneyball has snowballed across leagues.

The old-timer laments the loss of traditional football. Even millennials may long for the defensive wars of the early 2000s. But that era is gone, never to return.

Football fans of all ages and levels of sophistication have two choices: continue to resist, or join the 21st century.

At Jets X-Factor, we incorporate analytics into many of our columns. Michael Nania is our resident analytics expert. He blends the analytics with film study to provide a comprehensive breakdown of each play.

We decided to give New York Jets fans a primer on some of the more universally used analytics across the NFL. You will see these cited in many football articles or on NFL broadcasts.

This guide is not “the book” on NFL analytics. The field is still new in comparison to baseball’s sabermetrics. Baseball analytics has been well developed since the 1980s.

But these are some of the most accepted metrics, and therefore the most recognizable to readers.

Pro Football Focus grading system

In all sports, statisticians search for a single metric that best differentiates between players. Baseball data scientists have come up with Wins Above Replacement (WAR). WAR quantifies how many more wins the player added to his team above a replacement-level player.

Developing a catch-all statistic in football is much more complex. Baseball is a largely individual-based sport, making it easier to isolate each player’s contributions.

Football, though, involves 22 players on every single play. Rarely is the performance of one isolated from that of the other 21.

It is difficult to separate a running back’s production from the blocking of his offensive line. A receiver is limited by the accuracy and ball placement of his quarterback. A quarterback running for his life will have little opportunity to be effective.

A defensive lineman can have double-digit sacks while actually being mediocre as a pass rusher (2013 Calvin Pace) or have lower sack totals but do a solid job rushing the passer (Carl Lawson).

With their grading system, Pro Football Focus (PFF) attempts to assess how each player did his specific job on each play. Their analysts break down All-22 coaches’ film from each snap of each game, assessing each player on a grading scale of +2 to -2. The final grade is essentially the sum of the grades adjusted to a scale of 0-100.

Therefore, PFF grades provide an overall assessment of how each player performed compared to his responsibilities per play.

The issue with this system is its subjectivity. Who is grading each game? How well do they understand NFL offenses and defenses? How can they possibly know each player’s responsibility without having been in the huddle?

Jets X-Factor film analysts have found examples of plays that boosted a player’s PFF grade but were, in fact, poor plays in which the opponent bailed the player out.

If our analysts know this but PFF’s do not, then how can we trust that the grades are accurate?

This question becomes thorny when the PFF grade does not match the fan’s eye test. It is even starker when the grade is at odds with game film breakdown.

After all, PFF awarded Dan Feeney (75.9) and Nick Bawden (73.4) with better overall grades in 2021 than Elijah Moore (71.2) and Michael Carter (71.0). That certainly doesn’t match up with what was seen on the field by either Jets fans or film analysts.

However, PFF grades are still the most popular analytics for fans and media. People like numbers that are easy to compare, and PFF grades provide that. The most important thing is to avoid treating them as cold-hard facts due to their subjective nature.

Rankings/Qualifiers

Snap counts affect comparisons between players. Due to injuries or differing roles on the team, players’ snap counts vary. Therefore, analysts use different thresholds to compare and rank players.

“Qualifiers” or “qualified” refers to players who took a certain number of snaps. Analysts will typically use a certain threshold to ensure the data sample only includes players who played a significant number of reps (e.g., min. 40 attempts, 50 targets, 500 snaps, etc.).

ESPN’s QBR vs. traditional quarterback rating

Even old-school fans are familiar with and accept quarterback rating (sometimes referred to as “passer rating”) as a statistical tool. While the scale is wonky (why is 158.3 perfect?), the rating makes it fairly easy to evaluate quarterbacks.

Quarterback rating is more highly correlated with winning than just about any other NFL statistic, traditional or otherwise.

Still, there are inherent flaws in the quarterback rating metric. The biggest issue lies in comparing quarterbacks across eras.

In 2021, if Tom Brady had a better quarterback rating than Ben Roethlisberger, then Brady most likely outplayed Roethlisberger.

But if Brady had a 102.1 quarterback rating in 2021, is that better than Joe Namath’s 65.5 career rating? For that matter, is it better than the 92.6 mark Brady put up in 2004?

There are a variety of problems with traditional quarterback rating that cause the lack of translation across NFL eras.

ESPN’s QBR metric also assesses how well a quarterback played in each game. Though ESPN has never released the exact formula, we know that it is based on Expected Points Added (EPA).

EPA represents the average number of expected points that any given player adds or subtracts on any given play. The calculation is based on averages from the specific down-and-distance, field position, and game situation, among other factors.

EPA tends to reward running quarterbacks too much (which explains why Mitchell Trubisky was third in QBR in 2018) and punish sacks too heavily (which is why Joe Burrow was 12th in QBR last season).

QBR is also relatively highly correlated with wins. However, it is not necessarily a more effective statistic than quarterback rating despite its more detailed formula.

DVOA

This metric from Football Outsiders may not be as popular. Since we cite it fairly regularly on Jets X-Factor, though, it’s worth mentioning here.

Not all opponents are created equal. Running the ball well against the 2021 Jets is not the same as doing so against the 1985 Bears.

Therefore, DVOA adjusts statistics for the opponent. DVOA stands for Defense-adjusted Value over Average. However, it should really say “Opponent-adjusted Value over Average,” since DVOA measures both offensive and defensive efficiency.

DVOA takes into account field position, down-and-distance, score, time remaining, and other factors to evaluate teams and players. The values are adjusted to account for the differences in opponents.

DVOA is a more effective way to evaluate a team’s performance in specific facets of the game (i.e. rushing offense, passing defense, etc.) than raw yardage totals since it adds so many important factors into the mix. A team’s rankings in total yardage categories can be affected heavily by game flow (is the team usually trailing or leading?) and opponent quality. DVOA accounts for all of this.

The DVOA metric is calculated as a percentage. Since it yields a number over average, that percentage tells you how much above or below average the player or team’s performance is.

Because the defensive objective is to give up fewer points and yards than average while the offensive goal is the reverse, a good defensive DVOA will be as negative as possible, whereas a strong offensive DVOA will be as positive as possible. For example, the Bills had the best defensive DVOA of the 2021 season at -18.0% while the Buccaneers had the best offensive DVOA at +26.7%.

Average Depth of Target (aDOT)

Average depth of target (aDOT) refers to the average number of yards past the line of scrimmage at which a receiver is targeted. This number helps contextualize the production of wide receivers and quarterbacks.

Wide receivers who see deeper targets will generally yield lower catch rates than receivers who see shorter targets. A West Coast offense quarterback should have a higher completion percentage than one who heaves the ball downfield repeatedly.

With aDOT, we can get an idea of the style of football that each receiver and quarterback is playing. A higher aDOT average tells us that the player leans toward low-consistency/high-reward opportunities. A lower aDOT average is a sign of high-consistency/low-reward opportunities.

This metric is versatile. It helps evaluate cornerbacks and safeties, as well. It provides information about a team’s usage of a defender and the defender’s role within a scheme.

You will see aDOT frequently throughout the analytics articles on Jets X-Factor.

Big-Time Throw Rate (BTT%) and Turnover-Worthy Play Rate (TWP%)

These PFF numbers reflect their names. BTT% measures the rate of “big-time throws”. A big-time throw is “a pass with excellent ball location and timing, generally thrown further down the field and/or into a tighter window,” per the PFF website.

TWP% measures the rate of “turnover-worthy plays”, which are plays that could have reasonably become turnovers. It is a better QB statistic than turnovers, as it accounts for many turnover-related factors. TWP% punishes quarterbacks for dropped interceptions and near-misses while absolving them of the blame for dropped balls, tipped passes, and other unlucky plays that turned into INTs.

TWP% tells a more complete story than interception rates. This often explains the discrepancy between a quarterback’s turnover statistics and their game film.

Like PFF’s grades, these metrics are subjective, so keep that in mind.

The importance of rate statistics

Many standard leaderboards in football rely on aggregated numbers, such as total yards, total touchdowns, total tackles, etc. Those statistics can be misleading and are not useful for making fair comparisons. Players miss games due to injury, and even when healthy, not everyone gets the same amount of playing time or opportunities.

Therefore, it is important to compare players using unit rates.

For example, tackles are a notoriously useless statistic. They say nothing about how efficiently the player performed overall. We need to compare a player’s tackle total to his missed tackle total if we want a true gauge of how consistent he is at tackling.

Missed tackle rate is a crucial stat for context.

This is a big reason why linebackers like Quincy Williams may not be as effective as their tackle totals suggest. Williams racked up a career-high 110 tackles in 2021, ranking 23rd among linebackers, but he also missed 14 tackles, tying for 17th-most. His missed tackle rate was the 24th-worst out of 66 qualified linebackers (min. 500 snaps).

Rates are crucial in many aspects of the offensive side of the ball, too. With players across the league getting such differing numbers of opportunities to make plays, using unit rates puts everyone on the same plane.

Corey Davis‘s 2021 season gives us a good example of a player who performed much better on a per-play basis than his aggregated totals would lead people to believe.

Davis’ basic stats are not pretty: 34 catches for 492 yards and 4 touchdowns. But he missed eight games, left early in one of the nine games that he did play, and was part of an offense that did not allow its receivers to rack up many chances per game due to how quickly it was getting kicked off the field. These factors limited his total number of opportunities.

However, Davis’s rate stats are actually quite solid. Davis averaged 1.74 yards per route run, which ranked 35th out of 101 qualified wide receivers (min. 40 targets). He placed 68th among wide receivers with 492 receiving yards despite placing 89th in routes run with 282. Not too shabby.

Routes run

“Routes run” is referenced a lot here on Jets X-Factor, so it’s worth bringing up.

When we reference routes run, we are talking about the number of offensive snaps in which a player went out to catch a pass. Essentially, if they were on the field for a passing play and did not pass the ball or pass-block, it goes down as a route run.

This tells us the number of opportunities a player got to contribute as a receiver, helping us contextualize production. It compares everyone on the same plane regardless of their playing time.

For instance, Corey Davis averaged slightly fewer receiving yards per game than Adam Thielen (54.7 to 55.8), but Thielen got the chance to run more routes per game (31.3 to 34.3). That’s why Davis takes the edge in yards per route run, ranking 35th of 101 WRs with 1.74 yards per route run while Thielen places 47th of 101 with 1.63.

Pressures

We also use the “pressures” stat frequently at Jets X-Factor. It’s a great tool for measuring a player’s impact as a pass rusher.

As tracked by PFF, a player’s pressure total is a combination of three things: sacks, hits, and hurries. If a player has five sacks, five hits, and five hurries, he has 15 pressures. Simple as that.

On their own, sacks can be a misleading stat for evaluating pass rushers. For one, they only represent a very small portion of a player’s total impact; sacking a quarterback is not the only way to cause havoc as a pass rusher. Plus, there is luck involved with sacks, as so many things that are out of a player’s control need to go right for him to be presented with an opportunity to get one.

Pressure totals are a much more all-encompassing method for evaluating how frequently a player wins his battles as a pass rusher and creates problems for the quarterback.

From a rate perspective, “pressure rate” is something we often reference. Simply put, pressure rate is the percentage of a player’s pass-rush snaps (defensive snaps in which they were asked to rush the quarterback) in which they recorded a pressure.

Allowed pressures are also used for evaluating pass-blockers on offense. When referring to pressure rate for an offensive player, the stat tells us the percentage of the player’s pass-blocking snaps in which they were credited with allowing a pressure.

Expectation metrics

NFL Next Gen Stats (NGS) compiles many expectation metrics. This leads to the words “over expectation” in the statistics shown on NFL broadcasts each Sunday. NGS analysts provide a page of such numbers to the broadcast team before each game.

Expectation is a mathematical concept, not just an emotional one. It is calculated based on average results from similar situations. These metrics account for real-time player tracking data, time remaining, down and distance, score, and many more factors to determine the expected value of each play.

Expectation metrics include completion percentage over expectation (CPOE) and rush yards over expectation (RYOE). These numbers tell you how well a quarterback or running back performed compared to how an average player would perform in the same situation.

RYOE isolates a running back’s performance from his offensive line’s, since the calculation factors in the quality of the blocking on each play. CPOE describes a quarterback’s performance better than completion percentage, taking into account the difficulty of each throw.

Real-time player tracking data is instrumental in the expectation metrics calculated by Next Gen Stats. RYOE uses tracking to estimate blocking quality by analyzing the positioning of all blockers relative to all defenders. CPOE uses tracking to deduce the tightness of the window that the quarterback is throwing into (distance between the intended receiver and nearest defender).

These metrics are also incorporated on the defensive side of the ball. Most commonly, you will see CPOE used for individual defenders in coverage, telling us how frequently they allowed a completion on throws in their direction relative to expectation.

Opposing passers completed -11.9% of their attempts over expected when targeting D.J. Reed last season (4th in NFL).

Reed joins a Jets pass defense that allowed +3.2% of passes to be completed over expectation in 2021 (32nd in NFL). https://t.co/U7WL3nqbXk pic.twitter.com/xdH74KfLCu

— Next Gen Stats (@NextGenStats) March 15, 2022

Relative Athletic Score

Testing numbers from pre-draft drills (like the 40-yard dash) have been a part of draft analysis for decades. But there was never a way to bring all of those numbers together into one neat package.

Relative Athletic Score (RAS), which is calculated by Kent Lee Platte at ras.football, makes it easier for analysts to evaluate a player’s overall performance in pre-draft athletic testing. It combines all of a player’s pre-draft testing numbers to spit out one 0-to-10 rating that sums up his level of physical ability.

RAS evaluates how well a player performed in each drill relative to the historical averages at his position. It also accounts for a player’s height and weight. A forty-yard dash time of 4.40 seconds is more impressive for a 230-pound running back than it is for a 200-pound running back. RAS accounts for this.

Essentially, a player’s RAS is equal to his all-time percentile ranking among draft prospects at his position. Breece Hall has a RAS of 9.96, which means his cumulative testing performance is better than about 99.6% of running backs in the RAS database (which goes back to 1987).

While one would think that a 5.00 RAS would represent “average” athleticism for an NFL player (since it’s the midway point between 0 and 10), the reality is that the bar is much higher.

The RAS database includes countless prospects who never actually played in the NFL. Many of those prospects severely lacked athleticism, so the bar for an NFL-caliber player is higher than the average among all players in the RAS database.

To demonstrate this, the average RAS among players who were selected in the NFL draft from 2020 to 2021 was 7.55. That number is a more accurate baseline for NFL-average athleticism.

Final thoughts about analytics

Through the data explosion that has stormed the NFL, analytics gurus are not usually coaching hires. They’re typically not the general managers of NFL teams, either.

Rather, the analytics staff is one part of an NFL front office, the size of which varies across teams.

Even as executives with analytics backgrounds find themselves in the top spot in baseball, that has not become the trend in football.

The complexity of football precludes this from happening. Nothing can replace game film, strong schemes, and good fundamentals. The metrics complement all of those components. Ultimately, though, the game is won through a combination of plan and execution.

Perhaps the least known use of analytics is in salary cap allocation. Many are under the illusion that the salary cap is just a construct of the mind. However, it is anything but.

Rather, the salary cap drives roster decisions, contracts, drafting, and so much more. Many analytical tools help general managers structure contracts, decide whether to sign a player, and choose which positions are most important for the team.

This is an overlooked area of analytics but might be the most important in team building.

***
Whether you love the metrics or hate them, they’re a part of the modern NFL. I hope this guide has helped you understand the current numerical landscape and appreciate the goal of analytics.