Myles Garrett, T.J. Watt and Brock Purdy: Why The Data and the Film Agree
Two polarizing football discourses blew up, and they are useful case studies in cohering "analytics" and "film analysis." Why I'm skeptical of Brock Purdy and rank Myles Garrett ahead of T.J Watt.
The conversation surrounding modern football evaluation almost always touches on the nature of “film” versus “analytics” in assessing player quality and value. But as much as we emphasize the differences between the two, there’s a lot more that these two apparently divergent schools of thought agree on than disagree on.
The value of quarterbacks has increased regularly, more than the increase in cap space, over the years and we’ve seen a corresponding decrease in running back contract size. These were not a product of data analysts entering front offices – these changes have been happening for the past two decades.
If we wanted to look at complete datasets, we’re restricted to contract values from 2011 and later. That also offers a convenient cutoff point because of the uncapped 2010 season, which skews the values.
It should be noted, however, that the trend of running back salaries decreasing goes back further than 2011 – the top contracts we have data for from 1996, 1997, 1998 and 1999 showcase average values that produce a cap hit of 14.2 percent, 15.8 percent, 14.9 percent and 14.1 percent of the respective total salary cap allocation for those years.
Those are higher percentages than top five quarterbacks in recent years.
This trend towards agreeing with data analysis has expressed itself in other ways too, with increasing positional specialization, changes in passing rates and the deployment of more receivers and defensive backs. That said, there still has been a significant change in the way that players are evaluated in front offices and by the media.
The introduction to increasingly sophisticated – yet abstract – tools to evaluate player production and performance has invoked a larger public debate about how to break down and analyze players. This has led to a split in “film watchers” and “analytics nerds” about positional value and about the effectiveness of specific players.
There are two standout cases to me this past week that really drive the point home: Brock Purdy and the debate between T.J. Watt and Myles Garrett.
Brock Purdy and His Winning Ways
Brock Purdy is a Conundrum
In the context of evaluation – especially for the purposes of prediction – statistics provide us with a convenient and sometimes powerful proxy for talent. Knowing that a quarterback leads the league in adjusted net yards per attempt tells us that this passer is likely playing football at a high enough level to help lead his team to wins.
Last night, against the New York Giants, San Francisco 49ers quarterback Brock Purdy led his team to a 30-12 win while throwing for 310 yards on 37 attempts to generate a 111.3 passer rating and 8.72 adjusted net yards per attempt. It was the sixth-best performance any quarterback has put together this year in terms of expected points per play.
He also wasn’t playing well for most of the game. His most ardent defenders argued during the game that this was unrepresentative of his quality – until he threw two impressive touchdown passes after the half, erasing the memory of inaccurate throws and dropped picks throughout the first half.
Until his final touchdown throw – a dime to Deebo Samuel at the edge of the end zone – Purdy hadn’t completed a pass over 20 yards all game. That’s generally fine, but it fed into the larger perception that Purdy really hadn’t made many challenging throws throughout his short career as a quarterback.
Complicating this discussion is the fact that last year, Purdy was a very different quarterback during the regular season and the playoffs. I pointed this out right before the NFC Championship Game.
Purdy’s time to throw … matches the league average of 2.73 seconds. It’s not quite as systematic as Garoppolo, a quarterback with a snap-and-throw process that saw him get rid of the ball faster than almost anyone else in the NFL, but it’s much faster than his college tendency.
The issue is that Purdy naturally wants to make plays. He doesn’t have the arm strength of many of the league’s best scrambling playmakers, which means his throws demand more precision.
As the weeks have gone on, he’s become more audacious with his playmaking. In his two postseason games, Purdy’s average time to throw jumped up to 3.21 seconds from his 2.73-second regular-season average.
Purdy loves to scramble and he loves to create. He also has a gift for anticipation and in-structure play. When he abandons that gift, he puts himself into a lot of trouble – the near-picks he threw against the Giants were one thing, but his wild play in the first halves against Seattle and Dallas last year was almost bench-worthy.
So the Purdy discussion will always be complicated until he settles into a playing style or learns to integrate all of these playing styles contextually into his play. Against the Rams this year, Purdy seemed to play looser – with a higher degree of inaccurate throws, a willingness to push downfield and some capacity to invite pressure.
Brock Purdy’s Playstyle Invites Scrutiny
In front of a national audience, Purdy seemingly played like he did for most of his rookie season – short, high-percentage throws that resulted in enormous yards after the catch; Purdy saw 197 of his 310 passing yards come after the catch, a percentage of 63.5 percent.
For context, that would rank number one in the NFL over the course of a full season. Last year, Jimmy Garoppolo led the league with a YAC percentage of 59.4 percent and no other player was above 55 percent. Purdy’s game against the Giants was the fifth-most YAC-dependent game of the 2023 season thus far.
His quick time-to-throw and short depth of target raised eyebrows in part because that characterized his most of his regular season play from last year, but proponents were quick to point out that this was a product of an absurd blitz rate from the Giants – one of the highest blitz rates we’ve ever seen in a game – above 84 percent.
That’s worth some investigation. Using Pro-Football-Reference’s database, which goes back to 2017, there has only been one game with a higher blitz rate. It just so happened that it came from the same coordinator who called so many blitzes last night, Wink Martindale.
He had already coordinated a blitz-heavy gameplan against Kyle Shanahan on December 1st of 2019. That day, the Ravens blitzed Jimmy Garoppolo on 87 percent of his dropbacks. The Ravens won that game 20-17.
Garoppolo got rid of the ball quickly in that game and was fairly reliant on yards after the catch. That said, he’s always gotten rid of the ball quickly and relied on yards after the catch, regardless of blitzes.
The same data tells us that the relationship between blitz rate and time to throw is nearly nonexistent and slightly positive. The same is true of the relationship between blitz rate and average depth of target (slightly positive) as well as blitz rate and the percentage of yards that a quarterback has that comes after the catch.
Even when isolating just the games where the blitz rate was staggeringly high – above 50 percent – the average pass traveled through the air 0.10 yards further down the field in those games than in others.
When comparing blitzed throws to non-blitzed throws, the average distance in depth of target is just half a yard. Even when just looking at the most heavily blitzed games we have in the data, there just isn’t much to tell us that this meaningfully reduces time to throw or distance downfield on targets.
While Purdy has been more aggressive this year, this has resulted in a higher proportion of dangerous throws – but ones that haven’t yet been punished.
When he tried those throws against the Giants, he invited further disaster. He was again lucky that the defensive couldn’t capitalize.
There’s a reason Pro Football Focus did not produce a positive game grade for Purdy, ultimately giving him a 61.1 overall for his play. They noted two turnover-worthy plays despite an otherwise safe playing style.
The 49ers receiving group forced 13 missed tackles, suggesting that the yards-after-catch were not merely a product of well-placed throws to lead receivers into space but a product of immense individual effort as well. On the year, Samuel’s eight avoided tackles are more than the number of missed tackles 25 teams have combined across their receiving corps.
All of this is to say that Purdy’s production seems to be driven more by the efforts of his teammates as well as a good degree of luck. After all, he’s had four dropped interceptions.
Traits-Based Evaluation
Generally, the pro scouts in the NFL front office do not evaluate players using statistical measures. They use a constellation of factors, like anticipation, throwing motion, release time, ball velocity, accuracy and so forth.
That might be why, for example, NFL evaluators seem to be a little bit lower on Purdy than his ranking in adjusted net yards per attempt (fourth in 2022) or EPA per play (seventh). The Athletic’s survey of NFL coaches and executives before the season placed him in Tier 4, or about 24th.
Jeremy Fowler’s preseason survey of executives for ESPN did not feature Brock Purdy once – not in their top ten, honorable mention or “receiving votes” category. Many of the players Purdy lost out to are predictable, but he outproduced (at least statistically) most of them.
All in all, 19 quarterbacks received top ten votes in that exercise, placing them all ahead of Purdy. Stephen Ruiz ranks quarterbacks weekly for the Ringer using film analysis. Ahead of week 3, he felt Purdy fell short and ranked him 27th because of an aggressive risk-taking mentality without the arm or judgment to back it up. Ruiz also had pocket presence concerns. Derrik Klassen, ranking Purdy 19th for 33rd team, argued that Purdy just does not have the deep ball accuracy to be a consistent threat.
We’ll see what happens with Purdy’s career in the long run, especially if we find him in situations that require him to push the ball with a less impressive supporting cast or less capable playcaller.
For me, a team of strong tacklers willing to force short throws without vacating the middle of the field will likely produce a positive game plan against a Purdy-style passer.
That was my prediction against Philadelphia – he had only faced one good tackling defense in his magical run last year, the Dallas Cowboys. Only one other team had an average missed tackle rate – the Washington Commanders – and the rest had below-average marks in making their tackles. It’s unsurprising that his two lowest EPA-per-play outings after his debut were against Washington and Dallas.
The Space Between Traits and Stats
Often, the divergence between statistical and traits-based evaluations can lead to meaningful insights; Nick Foles and Josh McCown weren’t expected to continue their impressive levels of production after their stunning 2013 seasons and they didn’t. Case Keenum’s 2017 didn’t repeat itself and the NFL didn’t think it would. So too with Andy Dalton in 2015 and Ryan Fitzpatrick in 2018.
This is not the same thing as traits-based evaluation being superior; we’re more than familiar with draft and free agency busts every year – the Vikings know about that given free agency failures like John Carlson, Donovan McNabb, Alex Boone, Mike Remmers and Greg Jennings.
Still, there are statistical clues in the profiles of players like Purdy that give us context; Purdy struggles to throw deep, struggles on long-developing plays – especially under pressure – and needs play design to create YAC opportunities, limiting the types of plays that can be put in a gameplan. All of that can be measured in various ways that are represented in the data.
This might be why the analytical community also seems to be suspicious of Purdy despite exciting top-level numbers.
But primarily, they provide us with a solid foundation to make trait-based observations more than they are to find new insights.
While there’s still always pushback when a quarterback with low-level production gets ranked above quarterbacks with higher passing efficiency, there’s still some broad agreement among experts for the general skill level of a quarterback.
This statistics/traits problem extends itself into other positions, too.
Myles Garrett and His Win Rate Ways
Keep reading with a 7-day free trial
Subscribe to Wide Left to keep reading this post and get 7 days of free access to the full post archives.