Because of rules changes in the past couple of decades that have greatly facilitated the passing game, no position in the NFL is as important as quarterback. The days when a Trent Dilfer or a Brad Johnson could win a Super Bowl on the coattails of a suffocating defense are gone. These days, you either have an elite quarterback or you’re looking for one.
In the past two decades, 18 of 20 Super Bowls have been won by teams with a quarterback who is in the Hall of Fame, is expected to be, or at least has a chance to be. (The exceptions are Joe Flacco and Nick Foles.) About 50% of losing Super Bowl quarterbacks have been non-elite over that same period.
At the very top, arguments about who the best quarterbacks are exist mostly for entertainment purposes. You don’t need stats to tell you that Patrick Mahomes is great, you just have to watch him a few times. The same goes for Josh Allen, even though he hasn’t reached a Super Bowl yet. Throw in a few others such as Joe Burrow, Lamar Jackson, and Justin Herbert. The farther down the quarterback list you go, though, the higher the stakes become, because at some point the list of quarterbacks good enough to lead a team to a Super Bowl victory is effectively complete.
Are there 15 quarterbacks on that list? Probably not. How about 10? Maybe. The list might even be shorter. That is the question every general manager faces. Is Daniel Jones on the list? That is the $160 million question, and it’s not yet answered, as Jeremy Portnoy discussed in his breakdown of Jones’ career this week. Jeremy reports that Jones is ranked anywhere from No. 11 to No. 17 in recent media assessments.
Are there reliable metrics of quarterback play?
Many of us just use the eye test in deciding who the best quarterbacks are, but Jones is only one of many starters below those top few that people disagree, sometimes vehemently, about. It would be helpful if there were one or more objective measures that make the case for or against particular quarterbacks in a meaningful way.
Robby Greer (@greerreNFL), the developer of the nflelo app, has done an interesting study testing the utility of various metrics for quarterbacks. A quarterback’s No. 1 job is to win, so Greer looks at how much of the variance in margin of victory (MofV) can be explained by various quarterback metrics that are out there (margin rather than W-L record, partly because people have occasionally been known to pay attention to football games for reasons beyond just who wins). Here are his findings:
Football is a 22-person game, so no one player can control who wins and by how much. It’s a testament to the importance of quarterbacks in today’s NFL that you can find metrics for them that explain even 30-40% of the variance in MofV. There are a few.
Expected Points Added
The winner is Expected Points Added (EPA, 40% variance explained). This is not a surprise, because EPA tries to isolate the QB’s contribution to his team’s scoring, whether by passing or running; because a play that gets a first down or gets a team into scoring position carries more weight than plays of the same yardage that do not; and because meaningless “garbage time” yards and scores count less than those that occur when the game is up for grabs. Here are the top 2022 QBs sorted by EPA: 1. Patrick Mahomes; 2. Josh Allen; 3. Jalen Hurts; 4. Joe Burrow.
It’s hard to argue with those first few, though you might be surprised that Jimmy Garoppolo (No. 8) and Dak Prescott (No. 9) are also in the top 10.
No. 11 is Daniel Jones, knocking on the door of the top 10 but not there yet. Jones’ EPA was negative or very small in his first three seasons (though 2021 was biased by the games he missed due to injury), but it dramatically improved in 2022.
Adjusted net yards per attempt
Adjusted net yards per attempt (ANY/A, 35%) is an interesting statistic that gets little attention, but it has the second-highest predictability of MofV. The formula for ANY/A is
ANY/A = (passing yards + 20*(passing TDs) - 45*INTs - sack yards)/(pass attempts + sacks)
Greer feels that some of the success of ANY/A is because quarterbacks who are winning take fewer sacks and throw fewer INTs, so it is more a symptom of winning than a cause.
Mahomes finishes second in this metric (7.9) to Tua Tagovailoa (8.4), so is ANY/A as much a receiver stat as a quarterback stat? ANY/A is normalized by attempts, so quarterbacks who only played part of the season, like Garoppolo, Brock Purdy, and even Sam Darnold, make it into the top 10.
Daniel Jones was No. 28 in ANY/A at 5.9, but 21st among QBs with considerable playing time. My first reaction was that this must be the consequence of Jones becoming a risk-averse quarterback under the Joe Judge/Jason Garrett regime, and continuing that last season, after flinging it downfield more under Pat Shurmur as a rookie. But actually, Jones’ ANY/A for his four seasons is 5.4, 4.9, 5.6, and 5.9. If we’re looking for evidence of improvement in 2023, a big jump in ANY/A may be a good place to start.
Passer rating and QBR
The traditional NFL passer rating (33% variance explained) and ESPN’s QBR (32%), both widely used, rank fairly high, with passer rating being a bit higher. Purdy led the NFL in 2022 with a 107.3 passer rating; Mahomes was third at 105.2 (behind Tagovailoa) and Jones was 15th at 93.5. QBR, which we have discussed before, ranks Mahomes the clear No. 1, ahead of Josh Allen, with Jones showing strong at No. 7.
Jones’ passer rating declined from his rookie year to the Judge/Garrett years but improved modestly under Brian Daboll and Mike Kafka. His QBR was about the same in 2022 as in his rookie season, but better than in the Judge/Garrett years.
One thing to note about all these metrics is that Jalen Hurts ranks in the top 10 in every one of them. Whatever questions may exist about how he would do on a less talented team, Hurts’ production in 2022 was outstanding by any measure ... including the fact that he might have had a ring if his defense had made a stop in the final two minutes. And he has improved in almost every meaningful category over his three years in the NFL. The same is true for Tua Tagovailoa.
Did Hurts get better because he got Devonta Smith to throw to in 2021 and A.J. Brown in 2022? Did Tagovailoa get better because he got Jaylen Waddle in 2021 and Tyreek Hill in 2022? Of course. Part of being a great quarterback is successfully exploiting the weapons you are given, and both have done that. The next part is continuing to improve once you have the weapons.
There isn’t a foolproof way to isolate a quarterback’s performance from that of the team around him unless you start multiple quarterbacks in the same season. The San Francisco 49ers performed that experiment in 2022. It’s not perfect, since each QB faced different opponents and may have had different personnel available to them when they played. At first glance, though, we can say that Jimmy Garoppolo and Brock Purdy were pretty much the same quarterbacks in 2022, with eerily similar yards per attempt, passer rating, ANY/A, and EPA per dropback, while Trey Lance hasn’t gotten there yet, presumably because he’s played so little. Basically what you get when you play the 49ers is the Kyle Shanahan system.
Some stats do not appear to be great indicators when it comes to the role of quarterbacks in winning and margin of victory. The Pro Football Focus passing grade scores relatively low (23% variance explained) in that regard. That may not be a bad thing, since it tries (probably not completely successfully) to isolate the quality of the individual player’s performance on each play.
Of more interest are standard stats that are sometimes used to bury rather than praise a quarterback. As a rookie under Pat Shurmur, Jones threw 24 TD passes in 12 starts, a pretty good number, but he also had 12 interceptions and 19 fumbles. The narrative going into his second year was all about him having to cut down on turnovers, and in the Joe Judge/Jason Garrett era, he did that admirably (10 INTs/9 fumbles in 2020, 7/7 in 2021). Along the way, though, he lost the TD passes, too (11 and 10).
Greer’s chart shows that INT% is a terrible indicator of a quarterback’s contribution to MofV (13% of variance explained) and that the Judge/Garrett risk-averse approach was misguided. TD% is a better indicator, but not by much (22%). However, the metric TD% - INT% does a fairly good job, explaining 29.5%, so if we’re looking for a simple box score stat that anyone can understand that has some predictive ability, that’s the one. It’s not ideal, especially for a quarterback like Jones, whose impact as a runner is substantial, but it’s at least of some use. It will be interesting to see whether Jones can boost his TD% with better receivers and whether he can throw downfield more as he did often in his rookie year, while keeping the interceptions down.
Meanwhile, what to make of a quarterback like Jimmy Garoppolo? He came close to a Super Bowl ring too, yet he is widely dismissed these days as a franchise QB. What would Daboll and Kafka have done with him, or Andy Reid, or Mike McDaniel, or Sean Payton? More than what Shanahan was able to do (which was almost enough)?
And what of Jared Goff? He almost won a Super Bowl too, had it not been for Bill Belichick implementing a two-high safety defense to stifle him. He got pigeonholed as a QB that was a Sean McVay system creation, yet he led a prolific passing attack in Detroit last year with only one elite receiver (Amon-Ra St. Brown) and he finished high in every metric we have discussed above as having fairly good predictability. Is Goff borderline elite?
Maybe Goff is a good Rorschach test for how we view Jones. All we can ask of a quarterback is that he maximize what he has been given. Jones hasn’t had as much of a chance as Hurts and Tagovailoa thus far, not even as much of a chance as Goff, but he did show in 2022 that in a better system, he looks like a better quarterback. Now let’s see whether with some more potent weapons in 2023, his metrics take another step up.