Palazzolo’s Pitch: Grades, Stats, and Analytics

Steve Palazzolo discusses the role of PFF grades in today's NFL analytics landscape.

| 3 years ago
Pitch-WK01a

Palazzolo’s Pitch: Grades, Stats, and Analytics


Pitch-WK01aAs we head into the regular season, I just want to touch on the hot-button issue of “analytics” in the NFL. Much has been written on the subject and our own Sam Monson did a fantastic job explaining PFF’s place in that big picture, but it’s something I’d like to expound upon.

PFF founder Neil Hornsby perhaps put it best when he called us ‘Performance Analysts’. Contrary to what some people believe, we’re not creating algorithms from boxscore data and assigning grades to a player based on various formulae and, as we’ve stated many times, we’re certainly not scouts telling you about how a pass rusher’s lack of ankle flexion hinders his ability to sack the quarterback, or how his marital status shows that he’s mature enough to handle the rigors of an NFL season.

We’re somewhere in between those two extremes as we certainly provide a boatload of useful statistics, and rather than tell you how ankle flexion is keeping a pass rusher out of the backfield, we’ll tell you exactly how often and how efficiently he gets there through our grades and corresponding numbers.

Let’s dig a little deeper.

Grades vs. Stats

Perhaps we’ve been unclear about it in the past, but we strongly feel that our grades trump the stats we produce. There are literally thousands of stats that can be pulled from the PFF site, and some have great value, but the grades are still the closest representation of a player’s actual performance.

For instance, our coverage numbers are nice to have, but using them blindly is a big mistake. Even those numbers often rely heavily on a player’s teammates. An example:

•  A cornerback is beaten on a 10-yard slant and subsequently blocked out of the play by his own safety as the wide receiver takes it for an 80-yard touchdown. While this play is technically the cornerback’s coverage, and he’s likely downgraded for giving up the first down, his coverage numbers show 1-for-1 for 80 yards and a touchdown. That looks much worse than 1-for-1 for 10 yards, but he essentially did the exact same work.

The safety on the other hand, was not the primary coverage man on the play, but his play will surely call for a massive negative grade. He may not even be “targeted” by the end of the game, but his work will still be noted in his coverage grade.

There are numerous examples in the grades vs. stats debate, perhaps none more common than our own Khaled Elsayed’s disdain for the sack. It’s not that he doesn’t like sacks, it’s more that just one number is often used to determine a pass rusher’s usefulness even though sacks usually represent less than 4% of a player’s pass rushing snaps. There’s certainly a big difference in a player beating the left tackle in 1.8 seconds and securing the sack, as opposed to cleaning up after an indecisive quarterback holds the ball too long and drifts into a pass rusher’s arms. The stat sheet will look the same, even our own Pass Rush Productivity will be the same, but only our grades will differentiate between the two. There are even a number of instances where a pass rusher makes his presence felt beyond the typical sacks, hits, and hurries, and those are all accounted for in the player’s pass rushing grade.

Analytics in the NFL

In football, the term “analytics” can mean a number of things, but it seems that those who are against it or simply don’t understand it, come back with statements like:

1. “You can’t do Moneyball for football!”
2. “If stats are so important, why don’t you just look at yards and see who the best players are!?”
3. “Stats are for losers!”

Those short-sighted responses generally require these simple answers:

1. No kidding.
2. That’s right it’s either “No Stats” or “Yards.” Your only two options. This may be the silliest of all the anti-analytics arguments.
3. Perhaps you’re looking at the wrong stats.

So what is at the center of the analytical movement?

At PFF, we’re working closely with a number of coaching staffs and analytics departments, and it goes well beyond, “look at this one number that says Player X is better than Player Y.”

Some of the problems/solutions we’ve encountered this summer:

•  One offensive coordinator was keen to show his quarterback every comeback route that Peyton Manning’s threw last season (process of elimination tells me it wasn’t the Broncos’ OC). A couple clicks in the database, and those throws are highlighted with week, game and timestamp.

•  One defensive coordinator told us how our player participation data gave him a tell as to whether a certain wide receiver’s alignment would result in a run or a pass. He adjusted his play calls accordingly and attributed it as a key to victory that particular week.

•  One scouting department used our grades as an internal check for their own team scouts.

Does that sound like analytics? Not really. For the NFL, it just sounds like a Tuesday. But that’s part of our appeal to NFL teams as we have a detailed database of formation data, grades, and tendencies that essentially functions as a quality control coach, scouting department, and more for all 32 teams. And it’s not that we’re trying to eliminate quality control coaches and scouts from the NFL, but we’re certainly making their lives a lot easier so they can streamline their scouting process for their next opponents as well as enhance their player evaluations.

Value of Grades to NFL Teams: We Give you the “What,” You Provide the “Why”

As mentioned, you’re not going to hear any scouting jargon coming from us, but we’ll tell you how well a guy does his job. Anyone can tell you that Geno Atkins has elite first-step quickness and plays with great pad level, but what does that look like on the field? Well, it obviously translates to production as his +80.0 PFF Grade was more than twice as high as the next-best defensive tackle, Gerald McCoy’s +31.7. The funny part is McCoy probably has the exact same scouting report, and perhaps that can be translated as the two players being interchangeable, but we’ve been able to quantify the difference between them.

That’s not to say that scouts are useless – no, we’re helping them too. Again, the goal is to enhance, not necessarily replace. If you report on a player being really good at “bending the edge” but he never gets to the quarterback, does it matter? Or does it matter if a wide receiver is a “hands catcher” if he only drops five percent of the balls thrown his way?

Scouts always run into sample-size issues as they only see a player for a certain numbers of games. While a player’s physical attributes are likely easy to define over a four-game stretch, nothing paints the picture quite like a 1,000-snap sample. When a scout catches a player’s worst few games, but he happens to be teeming with green over at PFF, perhaps said player is worth a second look. We’re here to help.

Perhaps our grades don’t match up with a certain scouting reports and a team errs on the side of the scouts. That’s fine, but our information is there as a check and balance for a profession that often lacks accountability. As we often tell NFL teams, we’ll tell you how a guy performed, it’s up to you to figure out if it’s technique, athleticism, scheme, family life, or anything else that could affect on-field performance either positively or negatively.

We’ve never advocated blindly looking at our overall grade and calling it a day. But if you use all of the tools at your disposal, you’ll get a pretty good idea of a player’s skillset. PFF overall grade and its subsets (rush, pass, receiving, run stop pass rush, penalty, etc) are a pretty valuable tool to have in the toolbox, whether as an NFL talent evaluator or as a fan.

Analytics for Fans

One of the neat things about baseball sabermetrics is that general managers and fans can often look at the same numbers and use them to either make a better personnel decision or perhaps enhance the latest water-cooler baseball debate. In the NFL, it’s a much more cloudy process as current analytics are either boxscore-based and lack the context needed to make them viable for personnel decisions in such a complicated sport such as football, or they’re centered around the team rather than individual players and certainly are of little use to team-builders.

The point? Many of today’s most familiar analytics are great for fans, but not necessarily of value to NFL teams.

Predicting that a team’s offense will regress to the mean doesn’t do a whole lot of good for that particular team’s coaching staff or for their upcoming opponents. It does, however, provide great fodder for fans to discuss the true value of each NFL team, beyond just win-loss record.

Going back to the example above, using coverage numbers on a standalone basis with no context is dangerous, whether for teams or for fans. We take pride in our ability to add to that context through the grading system that takes everything into consideration.

“But you don’t know the play calls!” we often hear. While this is true, neither does the advanced scout trying to get head start on the next opponent or the pro scout evaluating a potential free agent. They still, however, need to make a determination on a player’s worth. As one former NFL defensive lineman told me, “if you watch enough plays, you have an idea of what you’re looking for.” We strongly believe that our grades, and the stats that supplement them, provide great value to both NFL teams and fans alike.

And speaking of fans….

If you’re not a subscriber, now’s the time to join. The regular season is upon us and the entire PFF team has been training all summer to provide the most in-depth, detailed, timely analysis in our six-year history. The same analysis that’s used by all 32 NFL teams in some capacity can be yours for only $0.08 a day. We hope you enjoy it!

 

Follow Steve on Twitter.

 

| Senior Analyst

Steve is a senior analyst at Pro Football Focus. His work has been featured on ESPN Insider, NBC Sports, and 120 Sports.

  • Ben Fitzgerald

    If you needed it explained to you that PFF grades aren’t analytics then…

  • JW_Redmon

    Thanks for the article, it will help me in the future when I have debates with fans who are anti stats, analytics and grades. Although it may be impossible to change some of their minds. You can lead a horse to water, but you can’t make him drink.

  • bz2

    Interesting you should say that you get access to individual plays with timestamps, because I signed up for the one month free trial of the 2008 stats, and they don’t show anything other than aggregate statistics per player and per game. Could you give me an idea of what you get when you do sign up for the premium stats?

  • http://www.scardraft.com/ Scott Carasik

    So when do these timestamps come into play. I’ve yet to get those and I’ve had premium for years.

  • Whizzer Wilkins

    I very much appreciate the difference in nature between PFF grades and stats. One thing I’ve always wondered, though, is how much subjectivity is involved. In the given CB/Safety situation, for example, judgment has to be used to apportion the blame between players — perhaps you could argue the CB gets a -0.5 for failing to prevent the 1st & the safety gets a -1.0 for blocking his own man, or perhaps it’s the other way around since if the CB hadn’t allowed the pass in the first place the TD never would’ve happened. In a game-grading situation where you’re going through ~100 snaps, grading every player, and presumably can’t spend too much time on each one, presumably there must be some quick judgments and potential for bias. Is this something that’s controlled for? Is there a set grading rubric/are games reviewed by multiple graders? What reassurance do we have that Atkins really does deserve to be so far ahead of the next guy, and it wasn’t just a case of the guy who graded most of the Bengals games being a massive D-line fan?