FMIA Guest: PFF On How Data Is Changing NFL’s Present And Future

Getty Images

Peter King is on vacation until July 15, and he lined up some guest writers to fill his Monday spot on Football Morning in America. Today, it’s a collaborative effort by Pro Football Focus, the popular football analytics website.

Previous guest columns: Fred Gaudelli (June 3) | Nick Hardwick (June 10)

By Pro Football Focus

As they famously said on the old Monty Python show, “And now for something completely different.”

Seven different voices representing PFF will handle the writing duties in various sections, including the site’s founder Neil Hornsby, who will provide a history lesson of football analytics in recent years and detail the changing attitudes towards them among NFL teams. As PFF’s majority owner for the past five years, ex-NFL star receiver and NBC Sunday Night Football analyst Cris Collinsworth will weigh in on how such analytical data is changing the game, from its impact on personnel and play-calling within the league to the fans who can’t seem to get enough data in their quest to succeed in either fantasy football and/or the newly legalized sports gambling scene.

Analyzing and putting football under the microscope is the task PFF tackles every week, and this week’s column offers it a new platform to share a few insights.

Changing the Game

How data is impacting the landscape of football

By Cris Collinsworth, PFF Majority Owner

When I broadcast my first NFL game during the 1989 season, I had absolutely no idea what to study or how to study. NBC provided me with a handful of newspaper articles, we watched some film at the team facility on Friday before the game, and we interviewed some players and coaches. I took notes, but I didn’t even have a board with the players’ names and numbers on it.

I was thrown into the deep end of the pool. This was going to be the shortest broadcasting career ever. Luckily, I had David Michaels as my producer (yes, Al Michaels’ brother). David had worked for years with Terry Bradshaw, and Terry had created these boards for calling games. The positions were aligned on this board where they would line up on the field. Offense facing defense, back-ups behind starters. All I had to do was fill in the blanks. Once again, my friend Terry was ahead of his time, and David showed me how to use it.

When I think back to those days, it’s pretty comical. Today, I could never read, watch or study all the data that I have available to me. In 2014, I bought controlling interest in Pro Football Focus. At the time we had 60 employees evaluating every player on every play of the NFL season. Now we have nearly 500 employees, providing data to 90 NFL and NCAA teams, multiple television networks and individuals who use it for private purposes. I won’t get into all the details, but if you are a data scientist, mathematician or IT specialist, and you love football, we are hiring.

PFF has already changed the way I think about building a team and play-calling. I can remember a time when everybody thought Andy Reid was crazy for passing 60% of the time. He doesn’t look so crazy now. I remember when running backs were thought to be the most valuable position; now they are considered the easiest to replace based on our WAR (wins above replacement) metric. I think it is fair to say that now very few NFL contracts are negotiated without PFF data being at the heart of the debate. The agent pitches all the positive data about the player, and the team is loaded with all the not so positive data. Some of those negotiating stories are pretty entertaining.

But as much as the data has changed broadcasting, it has changed the game of football even more. “Gut instincts” are no longer good enough. Decisions must be made based on the data. Every year at the NFL Scouting Combine in Indianapolis, PFF meets with nearly every team. I always laugh when somebody starts telling me about “old school” coaches in the league. I won’t mention names, but some of the “old school” coaches have recently blown me away with their knowledge of the data.

I sat in on a meeting with a team that had seven data scientists, mathematicians and IT specialists from Harvard, Stanford and M.I.T. all in the room, and they were loaded with questions that would make your head spin. Luckily, I had the people in the room with the answers. Of course, there are teams that are not quite that sophisticated, and it is getting more and more difficult for them to compete. The data arms race is very real, and it is widening the gap between those who engage, and those who don’t.

The fans are now engaged in the data arms race, as well. Fantasy football had always created a market for data, but since May 2018, when the Supreme Court ruled 6-3 against the Professional and Amateur Sports Protection Act that effectively barred state-authorized sports gambling, a new aspect in the arms race has opened to fans. Whether anyone likes it or not, there is no stopping state-sponsored gambling on football and other sports. We have seen a real spike in our consumer sales of our Edge and Elite products. Some fans just want to know more about their team, others want to be the smartest person at the water cooler, some want to dominate their fantasy league, and others still are writing their own gambling algorithms. Regardless, there is no going back now. Data has changed the game.

5 Rising NFL Players

10

Five players who could end up in the PFF 50 next year

By Sam Monson, Senior Analyst

One of my favorite aspects of PFF data and grading is how it can spot the obvious coming when it’s still some ways off on the horizon—getting ahead of the curve and identifying talent before it becomes self-evident. Every year there are players who excel in limited snaps before ultimately being handed a larger role and workload for their teams. When they continue dominating, we wonder how they were ever seen as anything other than superstars.

Case in point: When Joey Porter was a star and the sack leader for the Miami Dolphins back in 2009, coming off a 17.5-sack season, we at PFF were clamoring for his backup –- a former undrafted pass-rusher who had not long before been playing in Canada -– to get more snaps because he was generating pressure at a far greater rate than Porter. Cameron Wake ultimately went on to be one of the best pass rushers of the past decade and looked it from Day 1 if you were seeing beyond the box score numbers.

Such examples are everywhere, and each year it’s always an interesting exercise to take a look through the PFF grading and predict the players that could take that next step if they get the right opportunity. This past week we unveiled our PFF 50—a list of the best 50 players in football entering the season—but in this case let’s look a year from now and predict some players who could make that list in 2020.

Levi Wallace, CB, Buffalo Bills: If there’s a player with the backstory to rival Wake’s, it’s Wallace. With precisely zero scholarship offers coming out of high school, Wallace walked on at Alabama, and eventually earned a starting job. Then he had to do it all over again when he went undrafted before signing as a collegiate free agent with Buffalo. As a rookie in 2018, he earned the highest PFF grade of any first-year cornerback, along with the highest coverage grade, and wasn’t beaten for a catch longer than 29 yards all season. Though he played far fewer snaps than first-round selection Denzel Ward of Cleveland, Wallace looks like a potential star in the making if he’s given greater opportunity in year two.

Bills cornerback Levi Wallace. (Getty Images)

Vita Vea, DL, Tampa Bay Buccaneers: At the other end of the scale, you’ve got Vea, a player who went in the first round in 2018 but fell off the radar a little because he began the season injured, then took a little while to get going and ultimately didn’t produce the box score production people want to see. Vea ended up with only three sacks, but had 23 additional pressures as a pass-rusher, 17 of which came in the final six weeks of the season. Over that stretch of play, his overall PFF grade was 86.4, and he had a top-20 grade at his position, hinting at what’s to come.

Mackensie Alexander, CB, Minnesota Vikings: Changing positions in the NFL can be a significant adjustment, and sometimes it takes time. The Vikings drafted Alexander in 2016’s second round and moved him inside to the slot after he principally played outside at Clemson. His transition wasn’t smooth, but he has now seen his overall PFF grade improve each year of his NFL career: from 47.5 as a rookie, to 54.1 in 2017, climbing to 78.1 last year. Over the final half of the season, he was the highest-graded cornerback in the league at 88.2, surrendering just 80 receiving yards in a seven-game span. Alexander could emerge as a force with the right opportunity in 2019.

O.J. Howard, TE, Tampa Bay Buccaneers: Another former first-round pick, Howard has yet to top 600 receiving yards or 35 receptions in a season, even while tight ends are breaking receiving records across the NFL. Dive deeper into the numbers, however, and Howard looks primed for a huge season with an uptick in opportunity. His overall PFF grade last season was 89.4, higher than any other tight end outside of San Francisco standout George Kittle. And on a yards per route basis, he was third behind only Kittle and Kansas City star Travis Kelce. His average depth of target was 11.3 yards downfield, a top-five mark in the league, and now the vertical threat he brings is being linked up with new Bucs coach Bruce Arians and an offense that lives down the field.

Jon Halapio, C, New York Giants: The Giants are revamping their offensive line in a major way, but one of the unsung components of the rebuild is at center, where Halapio could emerge as a foundation piece to the new-look front. He began last year as New York’s starter before going down with an injury after just 116 snaps of action. But in those snaps, he didn’t allow a single pressure, despite almost 50 pass-blocking snaps against the Jaguars and their array of pass-rushing weapons. With vastly improved players beside him, Halapio could prove to be a significant upgrade as a player who isn’t being talked about much heading into 2019.

Collegians to Watch

What the data says about the 2020 draft

By Mike Renner, Lead Draft Analyst

When it comes to draft evaluations, prospects such as Quinnen Williams are an anomaly. For the most part, the sure-fire top-10 talents in any draft class have been dominating college football for multiple seasons before they declare for the draft. The 2020 class already has a handful of players doing just that. With elite PFF grades and exceptional athleticism, here are the names you can pencil into your top 10 for next April in Las Vegas:

Tua Tagovailoa, QB, Alabama: NFL offensive coordinators have been willing to build around mobile quarterbacks more and more in recent years, as they’ve realized how much pressure such passers can put on opposing defenses. Tagovailoa pairs exceptional athleticism and pocket presence with one of the most accurate arms in the country. He had the highest percentage of passes charted with perfect ball placement of any Power 5 quarterback last season and had the seventh-lowest percentage of passes deemed uncatchable. He also added four touchdowns with his legs on designed runs. That combination is going to give NFL defensive coordinators headaches.

Alabama wide receiver Jerry Jeudy. (Getty Images)

Jerry Jeudy, WR, Alabama: The Alabama offense is utterly stacked once again with Jeudy being the next big thing in the lineage of ‘Bama wide receivers, and he has a good chance to go higher than either Julio Jones (No. 6 overall in 2011) or Amari Cooper (No. 4 in 2015). That’s because there isn’t much the Biletnikoff award winner can’t do. Last season, he broke 17 tackles on only 68 catches, dropped a mere four passes, converted 7-of-11 contested opportunities and scored 14 times. Those are video game-like numbers as a true sophomore.

A.J. Epenesa, DE, Iowa: I’ll forgive you if you don’t recognize Epenesa’s name. Heck, he didn’t even start for Iowa last season. But if you go back and watch their games closely, you’ll see why he made this list. Listed at 6-6, 280 pounds, Epenesa moves like a man four inches shorter and 40 pounds lighter. On only 243 pass-rushing snaps last season, Epenesa recorded 10 sacks, 10 hits, and 26 hurries. That’s good enough for a 90.4 pass-rushing grade—the second highest of any returning Power-5 edge player in the nation.

Chase Young, DE, Ohio State: Young is the only guy Epenesa trailed in terms of pass-rushing grades for returning Power-5 edge rushers, and he seamlessly replaced Nick Bosa’s production along the Ohio State defensive line after the future second overall pick went down with core injury early in the season. Now it’s Young’s turn to be a top-5 pick in 2020. The sophomore led the nation with 75 pressures in 2019, and he rarely came off the field, seeing more snaps (783) than Bosa did in any single season of his college career. Listed at 6-5, 265 pounds, Young checks the size, athleticism, and production boxes in a big way.

Bryce Hall, CB, Virginia: It was a major upset when Hall decided he was returning for his senior season in Charlottesville. There’s a good chance the 6-1, 200-pound cornerback would have been the first defensive back off the board had he declared, because with Hall’s length, athleticism and ball production, he fits the mold of a modern NFL cornerback. Last season he notched an absurd 23 forced incompletions to lead the country. Maybe the biggest reason Hall decided to return though is because he’s capable of some jaw-dropping interception totals, underachieving last season in that department. He picked off only two passes in 2018, but led the nation with six dropped interceptions, two more than any other player in the FBS.

The PFF Mantras

20

PFF findings that are surprising, useful

By Eric Eager, Data Scientist

The opportunity to study football with math is one we try to approach with humility, because many results run contrary to what we’ve believed for generations and many findings are merely steps toward the truth instead of complete proof. We have a unique data set that includes PFF grading for every player on every play, and our first question, like the first question of most people, was whether the grading was actually useful.

PFF grading is not only far more stable (meaning consistent from season to season or week to week) than traditional metrics like passer rating or yards per carry, but is also more predictive should you care about figuring out who might win before the game is played. While many point to the inherent subjectivity of our grading, it’s important to realize not only are all stats subjective in some way, but that subjectivity doesn’t matter if they satisfy the two criteria mentioned above.

Here then are some of our findings that are the most surprising and useful to football fans and teams alike:

To predict quarterback play, look at how he performs when he’s in a clean pocket, not when he’s under pressure

Pressure is indeed important, and it affects every quarterback negatively. However, we find out very little about how good a quarterback is by looking at his pressured drop-backs. Indeed, we found that quarterback play (however you measure it) from a clean, un-pressured pocket tells you the most about quarterback play.

This makes some sense intuitively. Pressure happens in many ways—via the blitz (with compromised coverage), with a four-man rush (with good coverage on the back end) or some combination of the two. Pressure sometimes succeeds (often not the quarterback’s fault) and sometimes doesn’t. In this situation, you’re judging context more than you’re judging the quarterback.

It’s important to remember that quarterbacks spend about 70 percent of their drop-backs kept clean from pressure, which is why a quarterback like Ben Roethlisberger can be more than viable despite a passer rating under pressure below 70.0 in each of the past four seasons. By comparison, his passer rating in a clean pocket has been above 100.0 in each of those years.

Browns quarterback Baker Mayfield. (Getty Images)

From a clean pocket, we get to see the signal-caller get a fair shake, and we can ask questions like “Is he accurate enough?’’ or “Does he make sound decisions?” without muddying the surrounding circumstances. Effective play from clean pockets was a huge reason for PFF standing firmly on the side of Baker Mayfield’s potential in 2018 and Kyler Murray’s in 2019,  because we’ve gotten to see them succeed in those circumstances at the college level where others have failed.

Coverage is more important than pass rush, all else being equal

Analytics have a way of upsetting your sensibilities and upending some preconceived notions. Growing up, we’ve been conditioned to believe that pass rush is critical. And for good reason—pressure reduces passer rating substantially (by about 30 points) and nearly halves a team’s yards per play average. And we can see from the broadcast angle when pressure affects a quarterback. Because passing is so important, good pass-rushers have been the highest-paid members of most NFL’s defenses.

However, we found that not only does pass coverage (as measured by PFF grades) explain team success better than pass-rushing, but predicts it better as well. This helps explain why the winningest team in the league (New England) has used its only two big-name defensive free-agent signings on cornerbacks over the past decade (Stephon Gilmore and Darrelle Revis), and why defensive end Trey Flowers is currently a Detroit Lion. One need only to go back to last year’s playoffs to see how the quick passing game of today’s NFL mitigates even the strongest pass rush, with the Patriots racking up 78 points en route to the Super Bowl despite facing the vaunted pass rushers on the Chargers and Chiefs.

The caveat to this finding? As a trait, coverage tends to be less stable year to year. The upshot? Invest a lot into coverage, so that some subset of five or six of these players give you an elite group.

If you’re going to invest in a pass rusher, prioritize his pressure rate, not his sack rate 

Sacks are important and worth about two points for a defense. They usually end drives and often force turnovers. They are the ultimate goal of all pass rushers and rightly the most valuable non-turnover play for a defender.

But if you want to know how good a pass rusher is, and how good he’ll likely be the following season, look at his pressure rate. We at PFF define a pressure as a sack, a hit or a hurry on a quarterback, with the best pass rushers able to generate a pressure on 15 to 20 percent of their pass-rush snaps. For a pass rusher with 500 pass rushes during a season, you’re talking about 75 or so plays, versus just 10 sacks.

Generally speaking, “finishing” pressure with a sack isn’t really a trait a pass rusher possesses per se. If you want to predict a player or team’s sack total one year, use his or its pressure rates from the previous season. A 20 percent decline in a player’s sack total from 10 to eight, for example, is mostly noise. But a 20 percent dip in a player’s pressure rate from 50 to 40 is less so. Process over results.

History of Analytics

How NFL teams have changed their attitudes towards PFF and advanced data

By Neil Hornsby, PFF Founder

FMIA’s Peter King told me last week he estimated there were perhaps 400 people currently employed in analytical roles across major league baseball, then asked how many I felt there were in similar roles in the NFL? “Significantly fewer” was my instinctive response. His next question was the obvious “Why?”

I’ll tell you the same answer I told him. “It’s just been harder.”

Football was harder because, until very recently, you simply didn’t have a lot of data about the incredibly complex set of interactions that constitute a typical play. Baseball has always had a lot of data about its relatively simple, one-on-one encounters. It’s a little like trying to build a restaurant business with the use of only a couple of ingredients.

The lack of data in football was one of the main reasons PFF came into existence. Go back over the previous 10 years and tell me how well the left guard for the Tampa Bay Buccaneers was playing and how it compared to the rest of the league? Good luck.

Back then we were in a time when establishing the best middle linebacker was a matter of counting tackles, and an era in which people were trying to determine the best offensive line based on how many times its quarterback was sacked and how many yards the offense gained. By way of example, the Houston Texans last season gave up a league-worst 62 sacks. Everyone knows that. It’s a stat.

What you don’t know is players not on the offensive line were responsible for over a third of those sacks (21 were spread across tight ends, running backs, and even Deshaun Watson, who accounted for 14 himself). The Texans offensive line was not good, but Houston had far from the league’s worst pass protection (we had them ranked 20th).

Texans quarterback Deshaun Watson. (Getty Images)

Back in 2008, while a lot of people’s response to the lack of data was “Well, let’s work with what we’ve got,” our idea was slightly more ambitious. It was “Let’s try and collect the data we need.” In hindsight that plan seems more strategic than it really was. PFF was initially just a hobby based on collecting something that didn’t exist. The fact we now have all 32 NFL teams as customers, and by the time the 2019 season begins, over 60 NCAA clients, is still a matter of amazement to me. “How did that happen?” I sometimes catch myself wondering. We just got lucky I guess.

In that time we have certainly played a part in deepening the football data set. Since our inception we have collected over 13 years of NFL data (every game, every player, every play has always been our mantra), as well every NCAA FBS game played since 2014. In addition, a few years ago, the NFL brought in player tracking data (the x,y coordinates of every NFL player at 10th of a second intervals). Now that’s beginning to look more like it!

So now we have the data, so just throw a ton of data scientists at it and all the NFL’s analytical problems are immediately over, right? Not so fast. There are at least a couple of problems with that thinking.

1.) Any project manager worth their salt will tell you the time, cost, quality triangle they work within isn’t in perfect equilibrium. While you can add more resources to any project, you very quickly get into a law of diminishing returns when it comes to the idea that twice as many people means things get done twice as quickly.

Any vaguely competent business isn’t going to throw endless resources at a problem for which there isn’t even an estimated (never mind defined) rate of return. From what I have observed, most clubs’ response to analytics has been sensibly proportional to date and I’m sure they will add to that as the need arises. Sure, there are a few clubs off the pace and it will take a few years (and unfortunately maybe a change of general manager) before they catch up. However, this does at least account for some of the discrepancies between how baseball and football have embraced analytics.

2.) Although this will sound strange, the other major issue is finding someone with both the chops and skill set to introduce and use the data to effect change efficiently. Surely every major software company on the planet has a ton of incredibly smart people that can chew through such a (relatively) small data set in a few weeks? If only. I’m not going to embarrass anyone by naming names, but a few years ago one of the world’s foremost technology companies was given our data to work with. The “results,” delivered a few weeks later, had them proudly showing us how they had managed to count the total number of third downs in a season for the New York Giants. The graphs looked magnificent, however.

If a team ever wants its analytics department to be part of the crucial workflow within the football calendar (as opposed to a bunch of nerds with spreadsheets intermittently answering the odd tricky question), it will need at least one top-flight integrator to make it all hang together. In order to grow the department, you need to add meaningful and observable value, and finding a person who can conceive, articulate, plan and deliver this is not an insignificant challenge.

This role is hard to find in any business, but when you throw in the football component (someone with a deep enough understanding of the game that they can have a peer-to-peer conversation with a coach or GM), it becomes very difficult. Those individuals are slowly being identified and installed, but this is not a plug-and-play type of deal. In all probability, the team has compromised on one of those key attributes and may be trying to get perhaps a world-class business analyst up to speed on the “business of football.”

It takes time for a new function to gain acceptance in any business, much less one as set in its ways and traditions as football. For starters, just look at the uneasy alliance on some teams between the front office and the coaching staff and you begin to appreciate how hard it would be for a new group to succeed in that power structure. A “flywheel” methodology—the steady building of momentum predicated on multiple iterations of delivery, success and acceptance—is the only real way forward, and as that happens more and more people will be brought on board to build on that.

Football isn’t as advanced as baseball in analytics because it’s at a different point in its evolution, but it’s gaining traction. Let’s check back in five years and see where things are. I suspect we’ll not see anywhere near the disparity we do now.

Where Data Is Headed

30

How teams should be using PFF in the future

By Steve Palazzolo, Player Personnel Insights

Every NFL team is different. They are 32 ecosystems with 32 different leadership structures, communication patterns and decision-making styles. All 32 teams use PFF data to varying degrees. Just as there are 32 unique draft boards every April, each team has a completely different way of integrating PFF information into their day-to-day operations, from game planning, to creating a more efficient scouting process, to relying on PFF grades as part of player evaluation.

The player evaluation component has grown dramatically in recent years, as PFF data has developed from descriptive to predictive, and teams are just scratching the surface on how to best leverage the grades in every personnel decision. It took some time for PFF grades to gain traction as true descriptors of what was happening on the field, but many breakthroughs were made during our early years of existence.

For instance, we created a proper evaluation process for offensive linemen. We showed there are better ways to evaluate quarterbacks by isolating the timing, ball location and decision-making on every throw. We brought to light that pass rushers should be judged by far more than their sack totals, and we gave proper credit and blame to those run defenders who either blew up the play or got blocked out of it.

While PFF grades started influencing team decision-makers, player contracts, award-voting, fan perception and more, there’s still plenty of room to grow within NFL front offices. We’ve taken our descriptive grades and stats and spun them forward to better project what will happen on the field, or what should happen on the field.

The following questions are only a fraction of what can be answered within the PFF database as we continue to improve our team offerings:

Who are the best fits for each team, both in the draft and free agency?

The PFF grades have always been focused on the evaluation of each player, but now with PFF WAR (wins above replacement), teams have access to a valuation metric for each player. They can gauge how each move will shape their future, not just in an isolated manner, but league wide. One of PFF’s biggest advantages has been our ability to evaluate the entire league, not just one team, and that adds proper context for every decision a team will make.

PFF’s scheme information adds another important layer of detail in any decision. How well will a cornerback play in a zone-heavy scheme? Is a running back better suited for a gap or zone system? Which routes and route depths will maximize a wide receiver’s skill set?

Which parts of PFF’s data are most predictive of future performance at each position? What are the strengths and weaknesses for each player? Will they change over time?

Perhaps the most ground-breaking work will be our ability to redefine the “upside” of a player. Instead of just assuming that every player will improve, there is data that supports which parts of a player’s skill set will remain consistent over time, while other parts will fluctuate. The wise teams will know which pieces of the PFF data are most crucial when projecting future performance (think of baseball’s three true outcomes of home runs, walks, and strikeouts).

Which parts of a college quarterback’s performance best translate to pro ball?

With five years of FBS data, we can better project college players to the NFL. Especially at the quarterback position, we’ve been able to isolate the parts of college performance which best translate to the next level. Teams can use PFF data to project college signal callers like Kyler Murray or Dwayne Haskins based on the various situations they could see in pro ball. How would Murray fare in the same situation as Josh Rosen in Arizona with the same level of pressure and a similar play-calling mix? It’s all possible within the database.

How do I attack the opposing quarterback from a game-planning standpoint?

Coaches have been some of the biggest advocates of PFF through the years as our data has made their processes much more efficient, but they’re just scratching the surface on how they can further attack their opponents. They will continue to glean actionable insights from the data, more than just first down run/pass tendencies. Which quarterbacks struggle against disguised coverages and what are their tendencies when maneuvering the pocket?

Which coaches make the most optimal game-day decisions? When should I go for it on fourth down given my current roster and quarterback situation?

In-game decision making is another crucial piece of the equation and PFF data is already being used to optimize play-calling and fourth-down decisions. From studying situations ahead of time to evaluating play-calling tendencies in hindsight, this is an area that will continue to grow in the coming years.

Which coach or general manager is best for my team?

We’re just starting to properly evaluate coaches and quantify how much they can alter player performance or add value through play-calling. Front-office personnel, perhaps looking to justify a change, have already asked us where their current head coach ranked in these metrics.

Throughout the NFL, many teams have been using PFF to enhance their day-to-day processes while just skimming the surface of answering some of the organization’s most important questions using the data. Expect this to continue as we develop PFF IQ, our version of an NFL front-office decision-making tool. PFF IQ will touch every level of the organization as owners, general managers, and coaches will have all of the necessary data at their fingertips to help with every decision and ultimately aid in winning games and championships.

Outsider's Perspective

How a former player sees analytics

By Bruce Gradkowski, former NFL quarterback

As players, we spend each week preparing day in and day out to perform and try to win in the NFL. We condition our minds and bodies for the physical and mental grind of football. In the NFL, every week comes and goes rapidly and is spent preparing for a new defense or installing plays to put new wrinkles in the offense, all to gain the upper hand in six days or less. It’s all about maximizing time to get as much information to sink in as fast as possible. And when the week is done, it’s time to move on to the next opponent.

One of the things I learned during my time playing is there are so many people inside the organization that help us individually and collectively be successful.

The average career length of an NFL player is three and a half years. I was fortunate to last 11 years in the NFL, but it all has to come to an end at some point. After retirement, players are left trying to figure out what to do with their talents? The drive and determination that helped them make it to the highest level is now put on a bookshelf until they find their next passion. Gone is the focus that competition provides, as well as the familiarity of the locker room, the friendships, the bonds, and the ups and downs. Most importantly, all that preparation and film study? All done! What’s next!? And what business could possibly understand and relate to all of this?

So when I first sat and talked with Pro Football Focus, I was a little skeptical. How could people who aren’t in the locker room or haven’t played the position grade players on every play? What I’ve learned so far at PFF is there is so much of value to the NFL that comes from people outside of team organizations. The kind of folks who just love the game of football and want to help make it the best it possibly can be. They want to further the understanding of the game across the board and get as much detailed information about it out into the world as they can.

The thirst for knowledge and the desire to be great extends well beyond just coaches and players. PFF is a locker room in itself. It’s made up of a bunch of guys who are “gym rats.” True locker room guys! The owner is Cris Collinsworth, former star wide receiver. Cris is someone who understands what it’s like to be in a locker room. He understands what it’s like to watch film and break it down. I was very impressed with the knowledge throughout the company. Some of the employees might not have played the sport, but they do know it like the back of their hand.

The thing I admire most about PFF is the humbleness of continuing to try and learn and grow as a company day in and day out, just as if you were preparing for your weekly football game. There are coaches and former players in and out of the office on a weekly basis, helping to maximize the grading system, helping to “ultimate” the system all college and NFL teams have access to. (PFF’s flagship product for team customers is dubbed PFF Ultimate.)

When Cris Collinsworth asked me to draw some plays and defenses on a whiteboard for him, it made me realize how humble he is as well. For someone with so much stature and notoriety, he is always open to continuing to learn and grow. He is someone who has not only played the game, but has met with coaches and front office executives for years while broadcasting games, but is still humble enough to pick my brain.

For a bunch of guys who love football, working hard every day to give fans and teams what they want and need in the most detailed of ways is remarkable. NFL teams may have only a few assistant coaches breaking down film and coverages, but everybody gets involved at PFF. From Cris on down, it’s just a bunch of guys who love football and know how to work hard. I’m happy to say I’m a part of the PFF family now.

Gamblers and Data

40

How gamblers are using data to gain an edge

By George Chahrouri, Data Scientist

Our primary goal within PFF’s Research and Development is to continually get better at answering the question “What wins football games?” There are many aspects within that question, from college to pro player projection, draft-pick valuation and, of course, determining the value of different players, positions and facets. At the game level, the true test is making predictions for the eventual winning margin and hence the term “point spread.”

To get there, we start with analyses of different ways to measure overall team strength. Given the uniqueness of PFF grading, we can create metrics that go beyond the traditional use of the final score of a game to determine how well a team played and instead use mathematical techniques to create an “adjusted score,” getting rid of some of the flukiness that is rampant within the small sample size that is an NFL game.

Data analysis pioneers such as FiveThirtyEight have employed something called an “ELO rating’’ system that adjusts for the strength of opponent, using the final scores of games. We simply take it a step further by using the PFF grading system to give a different interpretation of the disparity between the teams in any given game.

PFF ELO isn’t the only metric that goes into our predictions, as there are nuances to each matchup that must be appropriately modeled. But just a few well-calibrated power metrics can explain about 80% of the spread that you see in Vegas or at your online book. Gambling on sports and football specifically tends to carry with it a slimy connotation (some of which is for legitimate reason), and our hope is to be a part of a movement that promotes a purer goal—to understand the study of football more soundly and to celebrate the mathematical techniques that we can all learn from.

With only 16 games per week and 32 teams playing, luck is a very real component of seriously gambling on sports. But having a model that can pick games anywhere near 55 percent correctly is an achievement, and puts an exclamation point on the importance of leveraging sound mathematical practices and pursuing the ultimate goal of truly understanding the game of football more each day.

10 Things I Think I Think

1. We think this is the year people begin, in significant numbers, to differentiate the decision from the result of the play. There is no way Frank Reich should have been criticized last year for deciding to go for it on fourth down in overtime against the Texans. If you want to criticize the sloppy execution go ahead but the decision to go for it was correct. If you have a dice that is weighted to come up as a six 50% of the time and every other number 10% of the time and you bet on something other than six, you are a fool regardless of the result.

2. We think an 18-game schedule with players only allowed to play in 16 of them would be fascinating. It should help with injuries—players won’t be playing in more games—and if they have a questionable injury early in the season, a team may be more likely to rest them to use one of the games they can’t play. It would also help prevent all-time records from being broken simply by players playing in more games.

3. We think the NFL is a “copycat”league, but the best teams and coaches are the innovators, not the copycats.

4. We think it matters far more how often you do something than if you can do something. In the ego-driven world of football the idea that a coach can turn a talented but poor player into something great is still prevalent. However, how many players in the last five years played poorly at the NCAA level and then turned it around in the NFL? It just rarely happens. Drafting on college production (within certain parameters) is the way forward.

5. We think right tackles are just as valuable as left tackles and should be paid as such.

Bucs tight end O.J. Howard and quarterback Jameis Winston. (Getty Images)

6. We think Jameis Winston will challenge for the passing yardage title in 2019. Last year he trailed only Josh Allen in average depth of target. These throws put Winston in a position to do great things at times (he was second among quarterbacks in the percentage of throws we grade as “positive”), as well as bad things (he was 21st in limiting negatively-graded throws). New Bucs head coach Bruce Arians has a track record of succeeding with high-variance quarterbacks like Winston.  In 2015 Carson Palmer had an MVP-caliber season under Arians, posting roughly the same average depth of target as Winston in 2018 and leading the league in percentage of positively-graded throws.  With Mike Evans, Chris Godwin and O.J. Howard a very capable trio of pass catchers, look for Winston to either make good on his 2015 draft position or give the Bucs no other option but to find his replacement the following year.

7. We think David Bakhtiari is the most underrated player in football. Over the past three seasons of PFF pass-blocking grades, Bakhtiari ranks No. 1 among tackles—by a distance! The No. 2 player on that list? Joe Thomas. Bakhtiari is the new gold standard of blindside pass protection, but doesn’t have the reputation or recognition yet.

8. We think the NFL now has a plethora of reasonable starting quarterbacks outside of the true superstars and any number of them can create top-10 production in any given year. Mid-tier quarterback production is more dependent on playmakers and scheme than ever before, so mid-tier quarterbacks need mid-tier contracts.

9. We think this is a big year for Sean McVay’s offense and, by extension, the teams that went on a binge of hiring away anybody who had ever been in contact with it to experience the same Midas touch. Late last year teams began to take away what McVay wants to do on offense, forcing the Rams to adjust, and Plan B never really looked like it existed, let alone was as effective. Successful systems have died quickly in the NFL in the past if they haven’t evolved, so 2019 is the year McVay needs to show evolution of his scheme.

10. We think we’ll miss the AAF terribly. It was high quality football and everyone on the football side responsible for putting a great product on the field deserves credit. With around 50 AAF players currently on NFL rosters there was obviously a real benefit to getting these players more experience.

Beernerdness

50

One of our IT developers, Alex Padgett, is also a talented home brew guy. He’s willing to take some big swings with each new thing—approaching it all as a challenge—and he’s churning out some excellent stuff. He’ll bring in a growler to share from his most recent batch every so often (a passionfruit / dragonfruit berliner weisse, a NE-style milkshake double IPA, and an imperial milk stout with coffee have been among the latest) and we are yet to taste a miss. No doubt in our mind that he’ll have something out for the world to taste at some point, but we are selfishly happy with the current limited access arrangement.

14 responses to “FMIA Guest: PFF On How Data Is Changing NFL’s Present And Future

  1. Are the PFF guys still here?
    In terms of the mechanics of playing the position of QB in the NFL, don’t the statistics show that Brees is ahead of Brady?

  2. To those bemoaning Wembley games since 2007 with “those dumb Brits don’t understand our football” type comments, it may be worth pointing out that PFF comes from the UK – in 2007. It was founded by Englishman Neil Hornsby who wasn’t happy with limitations of traditional NFL stats. In 2014 Chris Collinsworth became the major owner and PFF moved to Cincinnati.

  3. The problem with statistics in the NFL is the covariance matrix between players is significantly larger than in the NBA or MLB. Measuring that accurately either a baseball level of data that the NFL just doesn’t have access to given that it’s changing relatively rapidly or some clever thinking to get around. Coming up with a way to properly estimate coaching impacts, normalizing defense, and between players is what is always going to make NFL data much less accurate than the other sports for all but the very top players. I have much less faith in advanced NFL analytics than I do in a sport like MLB where the individual is on their own and their own production isn’t tightly correlated with anything outside themselves.

  4. PFF is in my opinion the most impacting organization that has changed football in the last 10 years. How everyday fans understand the game now using PFF’s analytics and writing from their staff is a 10X increase in the enjoyment and appreciation for the sport. I see PFF eventually being synonymous with being a fan of the game in general. Their continuing expansion into more meaningful and significant data will have a lasting positive impact on evolving this awesome game.

    We owe a debt of gratitude to the founder, Cris, Sam, and the whole gang that makes enjoying the science of football year round.

    When are we going to get a fully produced 1/2 hour PFF show sometime during the season on NBC?

    P.S. – Your live draft coverage was awesome.

  5. Great article! On of our big discussions while enjoying a local craft brew is the potential 18-game schedule. We came to a similar conclusion – limit the games played within the 18-games, but why not do it by quarters? Each player can play in 64 of the 72 available quarters throughout the season. Could be a fascinating study in strategy.

  6. So Collinsworth gets to pitch his own company on his Sunday night football broadcasts? That’s living the American Dream!

  7. Yeah, the 18 game schedule with players only eligible for 16 is brilliant. Teams will, obviously, sit their best players against teams that are sure wins. How will it feel to be a Jets fan when in all 18 games the opposing team rests its starting QB and skill players because….well…..you’re the Jets?

  8. Jamal adams not being In the top 5o players or even making ur top 5 list is crazy. He is top 10 best def player in league. What up southpaw yeah those terrible jets you obv have no clue

  9. I’ll say this: as a football traditionalist and an appreciator of modern-day observer of trends, I think that data analytics definitely has a place in football. Be it through educating the audience during a broadcast or defining a 53-man roster, good analytics is an effective tool that should not be ignored. Nevertheless, I become very wary of declarations about the efficacy of data determining future success. Each NFL play is, in and of itself, a microuniverse. It is faulty to draw conclusions about performance unless the data points are EXACTLY THE SAME. Measuring Baker Mayfield’s throws from a clean pocket from one game to another is a faulty metric of his performance because it does not take into account a myriad of variables (his health, weather conditions, field conditions, down and distance, field position, to name a few.) And as counterintuitive as it may seem, here is where traditional football instinct is a better way to evaluate performance, at least until we can account for every single variable. Let’s not forget that predictability is not why we watch (unless you’re a degenerate gambler); it’s the competition! The games themselves matter, and I fail to see the added value of including every variable without missing out on the grandeur of the game itself.

Leave a Reply

You must be logged in to leave a comment. Not a member? Register now!