clock menu more-arrow no yes

Filed under:

Wednesday Walkthroughs: Our favorite stats for understanding NFL football better

APC writers take a look at some of their go-to stats for evaluating players and teams.

Green Bay Packers v Minnesota Viki Photo by Jeff Gross/Getty Images

The nerds have won. Football is awash in advanced metrics, and even if it’s a little hard for the hardcore traditionalists to accept, it’s probably for the best.

Still, sorting through the avalanche of numbers can be a challenge. To cut through some of that clutter, here’s a look at the stats some of our writers here at Acme Packing Company turn to the most frequently.

Rcon14: DVOA

DVOA essentially seeks to measure the efficiency of either a player or team on a per-play basis, while adjusting for opponent. Given the fact that DVOA is relatively easy to contextualize with above 0% being above average and below 0% being below average, it doesn’t run into the complicated nature of trying to explain what a good or bad number is to someone who is new to the statistic.

DVOA also works for teams, units on teams, and players. Being able to use a stat that is actually useful across those lines is often difficult to find. DVOA also has a pretty significant number of seasons to go back and compare to, which makes it much more useful than other play-by-play stats, which are either difficult to get due to the programming required, or unavailable due to a lack of data.

Tex Western: ANY/A

Modern NFL football is first and foremost a passing game, and while DVOA is great — I use it heavily as well — I find that the best measure of passing success in a substantial sample size is Adjusted Net Yards per Attempt, or ANY/A. ANY/A is pretty simple, when you get down to it. It takes net passing yards (gross passing yardage minus sack yardage), performs adjustments for touchdowns (+20 yards) and interceptions (-45 yards), then divides it by the number of dropbacks allowed.

It’s an excellent tool for assessing an individual quarterback’s efficiency as well as a defense’s ability to defend against the pass, and in my opinion it is light-years ahead of the conventional passer rating. It also has some pretty nice even cutoffs for judging performance as well. It breaks down pretty well into the following rough :

  • Below 5: exceptional pass D or truly horrendous quarterbacking
  • 5-6: solid-to-good pass D or generally poor QB play
  • 6-6.5: average on both counts
  • 6.5-7: below-average pass D or above average QB play
  • 7-8: terrible pass D or very good QB play
  • 8-9: LOL-worthy D or exceptionally great QB play
  • 9+: literally only ever achieved by a QB in a full season four times in history

Let’s look at two case studies that help bear this out: Aaron Rodgers and the Packers’ pass defense. Both of these illustrate why this stat tracks exceptionally well with our perceptions of their play.

Rodgers’ 2011 season saw him post the second-highest ANY/A by a quarterback in history at 9.39, exceeded only by Peyton Manning’s absurd 9.78 in 2004. (Nick Foles’ 2013 and Matt Ryan’s 2016 are the other two over nine.) Rodgers’ other MVP year, in 2014, saw him at 8.65. He has five other seasons above 7: 2009, 2010, 2012, 2013, and 2016. However, the 2015 season and 2017 through 2019 were all below seven, with 2017 being his only year as a starter under six. That pretty well checks out with the general perceptions of his play over time.

The Packers’ pass defense over the last decade or so has been on both sides of the spectrum. The elite 2009 defense allowed a minuscule 4.2, while 2010’s team was even lower at 4.1. In 2011, that ballooned to 6.0, which would have been much higher if not for a league-leading 31 interceptions. More recently, the defense posted ANY/A against numbers of 7.1 and 7.7 in 2017 and 2018 before dropping that result to 5.7 in 2019.

Paul Noonan: DVOA

I make use of CPOE and EPA and all of the other newfangled stats, and I hate to be repetitive, but it’s hard to say anything other than DVOA. My only real criticism is that the name is a bit clunky, but it almost always passes the smell test and the eye test, it’s freely available, and it covers a very long period of time.

When I built out QBOPS and WROPS a few years back and took a critical look at the work I’d done, I concluded that I’d basically created a worse version of DVOA, which also provided some understanding of just how it operates. Nothing else balances success rate with explosion plays as well as DVOA. Nothing adjusts for the quality of defenses faced as well. Nothing adjusts for opponent quality as well. Nothing accounts for turnovers as well.

It’s ancient by the standards of advanced stats, but time has been good for it as Aaron Schatz has been able to iron out kinks. If you listen to the Off the Charts podcast, it’s clear why this is, as Aaron clearly understands the limitations of his own stat, and always works to provide context.

DVOA is simply great. If you look at the history of advanced baseball stats, you’ll quickly notice that many of the old ones no longer hold up. We never talk about VORP anymore, and the less said about defensive metrics (UZR for instance) the better. Football Outsiders had the benefit of that experience of course, but it’s still quite the accomplishment that one of the first “advanced stats” for football is still among the best.

Peter Bukowski: Total Points

My answer is DVOA, but for the sake of broadening horizons, I’ll throw out Total Points. What data often misses, Total Points attempts to capture. It’s the film grinder’s advanced metric. Developed by Sports Info & Solutions who pioneered some now-mainstream baseball stats, Total Points takes the grading of Pro Football Focus and the EPA of Football Outsiders and parses real value for plays by player.

A quarterback makes an accurate throw but the receiver drops it? The quarterback still created points with a good throw. They take the would-have-been value by historical EPA and assign what amounts to partial credit. A guard blows a block on a run play on 2nd down? Using historical EPA of negative runs on second down, that player receives the value of that mistake.

It requires specific, knowledgeable evaluation of the actual tape by people who have a clue what they’re looking at, blended with the analytic side. To me, it’s the kind of best of both worlds options we are missing from the NFL.

Jon Meerdink: Depth of Target

This is more a specific stat than a catch-all evaluator of overall production, but I think it’s really useful at capturing how players are used in ways that simple counting stats aren’t.

Depth of target simply measures how far downfield a player is when he’s targeted with a pass, but it reveals a lot about how teams are deploying their pass catchers. Davante Adams, for instance, had an average depth of target of exactly 10 yards in 2019, which is what you’d expect from a receiver who does his best work in the intermediate range. Speed burner Marquez Valdes-Scantling, though, had an average depth of target of 16.9 yards last year, indicating the Packers asked him to go long almost exclusively.

You can apply this same stat to other pass catchers, too. Aaron Jones and Jamaal Williams had fairly similar raw receiving numbers once you adjust for how many opportunities they got, but digging into the depth of target numbers, you can see that while Jones worked deeper downfield, Williams was almost exclusively targeted behind the line of scrimmage, indicating he was more of a checkdown and screen pass receiver than a true threat in the passing game.

Both Sports Info Solutions and Pro Football Reference offer depth of target stats, and you can compare their numbers to get a good picture of who’s doing what when the ball is in the air.