Happy football Friday everyone.
Week 4 is here, commenced with an impressive win by the upstart Detroit Lions, 34-20, over the Green Bay Packers. Week 3 brought us some performances for the history books, highlighted by a 70-point outburst by the Miami Dolphins. The Cleveland Browns continued their torrid pace, allowing -0.4 EPA per play, leading the league in EPA per pass play, and landing second in EPA per run play. The Cardinals notched their first win of the Jonathan Gannon era, while C.J. Stroud got on the board for the first time in Houston. Chicago, Denver, and Carolina stayed winless, as did the Vikings, who have now lost four straight one-possession games, after winning their first 11 such contests last year.
Using the data from the new and improved SumerSports.com, we will go over some of the thoughts and predictions I had a week ago, along with providing some for this coming week.
Week 4’s slate includes six divisional games, including a matchup in Orchard Park between the Miami Dolphins and the Buffalo Bills that could go a long way towards determining the first seed in the AFC. The four winless teams play each other, with a lot of soul searching for the two that lose. Let’s dig in.
DET @ GB Review
Detroit, favored for the first time in Lambeau Field against the Packers’ starting quarterback since 1986, dominated the Green Bay Packers in the first half en route to a 34-20 win. The Lions now have sole possession of first place in the NFC North for the first time since Week 2 of 2017.
After an early interception that the Packers turned into a field goal, Jared Goff managed the game, throwing for 210 yards and a touchdown, including over 50 yards to Amon-Ra St. Brown, Sam LaPorta, and Josh Reynolds. David Montgomery shouldered the load on the ground with 32 carries for 121 yards and three touchdowns.
A Trend I’m Monitoring
One of the things we discussed last week was NFL defenses optimizing for limiting big plays, even if that came at the expense of being able to get a three-and-out when trailing. One of the reasons for this dynamic is the two-high safety looks that continue to spread across the league.
In the early-to-mid 2000s, the rise of the Tampa 2 defense, named after the legendary Monte Kiffin’s approach to defense, put less stress on cornerbacks, preferred smaller linebackers, and required pressure without a blitz. These defenses, not wholly unlike the current approaches, did quite well, with the Bucs themselves winning a Super Bowl in 2002, followed by the Colts in 2006 – who triumphed over a Bears team that deployed a similar scheme under Lovie Smith. The Minnesota Vikings defenses from 2006-2008 were historic against the run, and after they got Jared Allen in 2008, devastating against the pass as well.
Why did the Tampa 2 go away? One possible explanation that has been given is the rise of the tight end, who if athletic enough to stress the seams can neutralize a lot of the benefits of the Tampa 2. Tony Gonzalez, Jason Witten, Antonio Gates, Jeremy Shockey, Todd Heap, Alge Crumpler, Dallas Clark, Randy McMichael, Chris Cooley, Desmond Clark, Marcus Pollard, Kellen Winslow, and Heath Miller all came of age during that time, while Shannon Sharpe hung on enough to catch passes for over 3,000 yards during the 2000s.
To contrast, of the top 12 tight ends during the last decade (2013-2022) in terms of receiving yards, only Travis Kelce, George Kittle, and Mark Andrews are still playing at anywhere close to the peak, with guys like Zach Ertz, Rob Gronkowski, Jimmy Graham, Greg Olsen, and others either retired or playing significantly reduced roles. No tight end in the league this season has a 100-yard receiving game, with only one player having earned more than 20 catches through three weeks (the below table is from before Thursday Night Football).
This led me to a question on the SumerSports Show with myself and Sumer CEO Thomas Dimitroff: should college programs that are rarely in contention for national championships go out of their way to develop talent at positions like tight end? Notice above that three of the four most productive players in the league at the position are from Iowa, which runs an offense that is near the bottom in terms of efficiency in the NCAA each year, despite harboring talent at this position. Tight end is both a position that doesn’t move the needle as much at the college level and one that is often cited as a difficult position in which to develop, as the player has to learn the nuances of both route running and blocking at the NFL level.
Without such development hubs, it will be interesting to see if there will emerge the kinds of talent requisite of beating shell coverages out of existence the way tight ends pushed the Tampa 2 out of being the league’s premiere defense.
Something I’m Buying
Another thing Thomas and I talked about on the recent SumerSports Show was which metric, if we could only pick one, we’d use to project quarterback play from one situation to the next – whether that be college to pro or NFL team to NFL team. I sent out this Twitter poll, which split the audience evenly:
they are both important, but if you had to rank them, which one comes first:
— Eric Eager 📊🏈 (@ericeager_) September 27, 2023
While there are several traits one needs to play the quarterback position, and they all matter, if I had to choose one, it would be pressure/sack avoidance. As I wrote in 2019, the quarterback controls a great deal of their pressure and sack rates, with the former having a year-to-year correlation that dwarfs that of every other quarterback metric created.
One of the reasons for this might be that pressure, and especially sacks, are easier to measure than something like accuracy. Raw completion percentage is rife with the biases inherent in how a quarterback plays. Sure, Sam Bradford set the NFL’s single-season completion percentage record in 2016, but he did so with an average depth of target of just 6.4 yards – which was one of the worst marks in football that year.
Completion percentage over expected, which derives a great deal of its signal from depth of target, is an improvement, but there are still things a quarterback does that engineers the expectation of the play, especially when things like separation are put into the model (I get into some of that in my new book). Improvements like charting accuracy instead of completion percentage have improved things, but still not to the point where such metrics have the same predictive power as pressure/sack avoidance.
My favorite anecdote for this argument is the career of Dan Marino. The Dolphins legend, who has a case for being the best quarterback of all time, never once led the NFL in completion percentage but led the NFL in lowest sack percentage 10 times, including the first seven years of his career. This was arguably the best seven-year stretch to start a career in NFL history:
Something I’m Selling
I’m selling the idea that the NFL needs to do something about the Eagles’ famed “tush push”; the play in which Philadelphia runs a quarterback sneak and the players behind him push him forward. Firstly, as we saw with Patrick Mahomes in 2019 (and we see every time Andy Reid uses literally anyone else to run a sneak, including former Oklahoma quarterback and current third-string tight end Blake Bell) there is injury risk to running that type of play, so while the play is difficult to defend, it is not without risk.
Secondly, the inability to stop something initially should not lead to banning such a play. As stated above, when teams struggled to stop deep passes to players like Tyreek Hill from Patrick Mahomes, they adjusted (and then Kansas City adjusted, alas).
Last year, NFL teams went for an average of 9.5 fourth down and one yard to go situations during the regular season, converting about two-thirds of them. The average conversion of a fourth and one netted about two expected points added (versus losing three-and-a-half points with a failure).
It’s hard to know how much more a team would go fourth down and one situations if they set out to deploy the tush push, but for reference, the Eagles went for 14 such plays last year during the regular season. Splitting the difference by about a third, adding one-and-a-half more fourth and one attempts per season, and increasing the success rate on such plays to a little under 90% (which is probably high), would mean a team adds about three additional fourth-and-one conversions to their season.
This nets out to roughly 17 additional expected points in a season, roughly one point per game, which is about half of a win each year. That’s actually a decent sized number. However, this assumes that no other team deploys the play, and that efficiency rates don’t regress towards normal averages. Thus, it might be reasonable to assume a number closer to a quarter of a win, which is not insignificant but also nothing from which to change the rules.
Shout out to @NFLPinnacleBeater on Twitter for this question:
How many expected wins would a well implemented tush-push add to a team on a season Eric?
— NFLPinnacleBeater (@NFLPinnacleBeat) September 27, 2023
My initial estimate was low.