For readers that have followed my work over the past few years, one thing should have been abundantly clear by now – I do not trust statistical analytics for football. Numbers are absolute, they are unwavering, and they tend to give off an air of black and white in a game that is almost entirely gray. Friendly debates among fans about who is the best "this" or who is the worst "that" have become inundated with buzz words like "yards per attempt" and "receptions per coverage snap" that are, for all intents and purposes, pretty useless.
Take yards per attempt, which is a statistic meant to reward efficiency for quarterbacks who may gain less total passing yards but tend to be more successful on a per-pass basis; the higher the average yards gained per attempt, the better. During one of Matt Schaub’s best years in 2009, his yards per attempt was a very respectable 8.2 due to the higher proportion of deep play-action passes thrown in Gary Kubiak’s run-first offense. That same year, Peyton Manning carried his Colts team to the Super Bowl with his slightly above-average 7.9 yards per attempt. What the stat does not say, however, was that the Indianapolis offense that was content to throw five yard passes all day to move the ball rather than trust their very, very weak run game.
Statistically, were Schaub’s numbers (4,770 YDS, 29 TDS, 15 INT, 8.2 YPA) better than Manning’s (4,500 YDS, 33 TDs, 16 INT, 7.9 YPA)? Yes, you could say that, but would anyone in their right mind assert that Schaub was a better quarterback than Manning? No, probably not. Two seasons later, both of these quarterbacks would go down for the year with injuries. The Texans would fight their way into the playoffs behind a third string rookie quarterback, while the Colts floundered without Manning into the first overall pick of the 2012 NFL Draft.
A similar argument still rages today among football fans about Manning’s replacement, Andrew Luck, and perhaps his most talented contemporary, Russell Wilson. Wilson supporters would argue that his efficiency-based statistics during his 2013 Super Bowl run, namely yards per attempt (8.2), make him the superior quarterback of the two. Luck supporters, however, generally lean on the fact that the Colts roster had to be damn near dragged kicking and screaming into the playoffs by one man, and that the eye test should reign supreme above all things.
If you had not guessed, I am in the latter camp, yards per attempt be damned. I do not care that Luck’s YPA (6.7) was vastly inferior to Wilson’s. I care that when down four possessions going into halftime against the Chiefs in the playoffs, Luck came back and willed his team to one of the most improbable victories in the history of the sport. Were the numbers pretty? Absolutely not, but the tape revealed a strong, poised, deadly quarterback who can get it done no matter the odds. One year later, Luck’s supporting cast has improved while Wilson’s has regressed. Take one guess as to which quarterback is still firing on all cylinders, and which one is trying to fight for a Wild Card spot.
Numbers, grades, and analytics can be construed to mean almost anything. Film study, however, can never lie to anyone if they know what to look for. That philosophy is in part why I routinely take issue with one of the most popular football analytics websites out there, ProFootballFocus.com. PFF, as many call it, not only compiles all of the raw traditional numbers that everyone knows and loves, but it has its very own "grading" system that aims to assign numeric value to individual performances on a snap-by-snap basis. The concept sounds fantastic, but in actuality it has become quite dangerous to modern football fandom. Connoisseurs of the sport have begun using these grades in arguments as traditional irrefutable statistics rather than taking them for what they are – a subjective and often misplaced attempt at trying to quantify a sport that is entirely qualitative. That is to say PFF has made their bones trying to give definitive numeric value to something that is so unbelievably complex that it cannot possibly be distilled into one single figure.
Take Andre Hal’s recent performance against the Philadelphia Eagles as an example. PFF graded out the Texans rookie at a very poor -5.1 on the day, with -2.1 of that coming from coverage snaps alone. For reference, anything within -1 or +1 on the scale is considered average, while outliers to that range are considered good or poor. The bigger the outlier, the better or worse the grade. J.J. Watt is generally in the +6 to +9 range, so a -5.1 for Hal is considered to be fairly atrocious.
However, when looking at the actual film of the game, a different picture takes shape. PFF credited Hal with giving up four receptions for 76 yards and a touchdown. Upon further inspection of the film, those numbers are given a completely different context. One of those catches was a "smoke route" from Josh Huff, who took it for 9 yards after the catch. Smoke routes are about as automatic as they come for catches and short yardage gains against off coverage, so is it really fair to fault Hal for that statistically? I would not say so. Another catch credited to Hal was a 4 yard gain to Zach Ertz underneath zone coverage that Hal wrapped up and stopped short of a conversion on third down. I struggle to understand how in the world a four yard gain that is stopped short of a first down is considered a negative play, yet technically it is when looking at pure statistics.
The other two catches for 63 yards are where the majority of Hal’s poor coverage grade comes from, yet he is not completely at fault for both of them. Take a look at this long grab by Jeremy Maclin on Mark Sanchez’s first snap of the game.
The Texans are lined up in their base 3-4 rather than a nickel front despite Philadelphia coming out in a three wide receiver set. Houston’s defensive alignment is rather ambiguous, suggesting it could be anything from Cover-Two, to Cover-Four, to even "Blitz-Zero" with both safeties covering the slot receiver and tight end while the entire front seven gets after the quarterback. However, Hal’s slight leverage to the outside of Maclin, illustrated by the red line, suggests that he is assigned to a deep quarter zone. Either way, he is clearly staking his claim on any fades thrown to his sideline while relying on strong safety D.J. Swearinger to help bracket Jeremy Maclin deep and inside.
Sanchez fakes the hand off to LeSean McCoy while Zach Ertz moves across the line of scrimmage to "block" back side. At this point Swearinger is reading the run all the way rather than executing his primary responsibility, which is making sure that Jeremy Maclin cannot do Jeremy Maclin things. Considering that the entire reason the Texans are in their heavier base front seven here is likely to allow the secondary to sell out against the pass, this is a terrible decision by Swearinger from the start.
Maclin and Hal are running full speed, yet Swearinger is still reading the fake.
By the time Swearinger finally does peel his eyes off the backfield, Maclin is already practically even with his depth.
And as the old saying goes, if he’s even, he’s leavin’.
If that play looked familiar to you, it is because the Eagles ran that exact same concept in the first quarter for a huge touchdown. PFF thankfully correctly credited that touchdown to Swearinger rather than Hal because of his abandonment of deep inside bracket responsibilities (again).
Just a few plays after Maclin’s second big catch of the day, Hal gave up the only reception of the game that I can truly 100% fault him for, this touchdown pass to his old teammate, Jordan Matthews. Hal was in great position to break up the pass, but like many young corners, he forgot the most important part of the play – actually turning his head around to see the ball. Had Hal had flipped his head the moment he saw Matthews look for the pass, it is possible that he would have prevented the score and forced a field goal attempt.
So when all is said and done, how much did Andre Hal really give up against the Eagles? When taking out the automatic catch and yardage on Huff’s smoke route, the four yard Ertz reception that got stopped short of the first down marker, and the 52 yard bomb to Maclin that was caused by D.J. Swearinger missing his assignment for the second time, Hal was truly "beaten" for just one catch for 11 yards and the touchdown. Considering that Richard Sherman was roasted by Odell Beckham more times this past Sunday than Hal was beaten all day against the Eagles, did Sherman really deserve a better coverage grade (-0.8) than Hal (-2.1)? I would not say so.
Giving up touchdowns is bad, of course, but Hal did a whole hell of a lot of good against the Eagles as well (including two dropped interceptions that could have turned the game around). One play never tells the entire story of a game, and thus that one play should never be allowed to weigh so heavily on any sort of objective, numerical evaluation of a player’s performance. I suppose that when looking at football analytics, fans and analysts alike tend to look at things in a vacuum. People are quick to point out how giving up a touchdown is obviously indicative of some kind of disastrous performance, yet many fail to recognize that same player having great coverage later on in the game that forces a quarterback to hesitate and get plowed by J.J. Watt for taking too long to make a decision.
PFF is not going away anytime soon, nor do I think it should go away. It is a very useful site that I subscribe to myself just to get access to raw data like snap counts, missed tackles, or defensive stops. I simply believe that football fans everywhere need to start pumping the breaks on taking PFF grades as gospel.
Football is not something that can be reduced to simple numbers and formulas. It is a living, breathing game that is constantly evolving and challenging our perceptions. Every game has dozens of snaps, and within just one of those snaps is a collection of dozens of stories. Both sidelines make calls based on their entire life experiences as coaches in these very specific scenarios. Captains on both sides of the field react and adjust to each other with moves and counter-moves based on their own knowledge and instincts. An offensive lineman may have been rolled up on in the last series, and his movement with his right ankle may be labored. Perhaps one of the defensive linemen notices this and attempts to exploit that weakness for a sack, or maybe he does not see the slight limp at all and the pocket stays clean enough for the quarterback to launch a long pass to the end zone. The wide receiver may haul that pass in, or perhaps he hears the safety’s footsteps barreling down towards his chest and drops the ball in a moment of lost focus. Hundreds of these abstract, qualitative factors play into every snap of every game. Trying to convert those intangible factors into a single number, at least from where I stand, seems like a futile endeavor.
Context is everything, my friends. When in doubt, I implore football enthusiasts everywhere to trust their eyes before trusting the numbers. We will all be better fans for it.
Eagles vs Texans coverage