2021 Passing Plus/Minus: A Tale of Two Kyler Murrays
NFL Offseason - Our annual review of the passing game continues with our look at 2021 plus/minus stats, where the changing of the guard continues. Drew Brees dominated completion efficiency stats in the 2010s, with only Philip Rivers and Matt Ryan posting even 50% of Brees' plus/minus totals in the decade. Well, Brees and Rivers are gone now, and Ryan is well on the back side of his career, so we're looking for the Drew Brees of the 2020s—the passer who will lead the league in completions over expectation. You might be tempted to look at the top of the DYAR leaderboards and assume that we'll be celebrating Patrick Mahomes, or Josh Allen now that he's past his rookie struggles, or the last hurrahs of great players such as Tom Brady or Aaron Rodgers. No, however. If you want to find the current gold standard for accuracy in the 2020s, you have to look down just a little bit, both on the DYAR tables and, you know, generally. There, you'll find Kyler Murray, 2022's plus/minus leader.
Passing plus/minus is a stat we annually track to help provide context to completion percentage. Given the location of a quarterback's passes, it compares his completion percentage in each area to historical baselines. This stat does not consider passes listed as "Thrown Away," "Tipped at Line," "Miscommunication," or "Quarterback Hit in Motion" by Sports Info Solutions charting. Metrics are based on how often a pass is completed based on the pass distance, the distance required for a first down, and whether the ball was thrown to the left, middle, or right side of the field. This is a counting stat, so more attempts are obviously a great thing for the purposes of what we're talking about here. Our Completion Percentage Over Expectation (CPOE) numbers may differ from other models around the Internet.
So far, Murray is leading the 2020s. He has taken a step forward each season, seeing his CPOE rise 0.5% to 3.9% to 6.3% as he has developed and grown in Kliff Kingbury's offense. Murray was second behind only Joe Burrow in terms of on-target passes last season, and that's not solely because of a diet of short passes. Murray wasn't just taking the empty calories of ten thousand dumpoffs to James Conner and shooting up the leaderboards; his connections with Christian Kirk and DeAndre Hopkins ranked third and sixth in plus/minus for quarterback-receiver pairs with an average depth of target of at least 10.0 yards. Yes, it looks like Kyler Murray will be the Drew Brees of the 2020s, and we should all be prepared for his upcoming dominance.
Well, no. It's more complicated than that, so let's dig into the data a bit. Here's where last year's quarterbacks stood, ranking all 34 passers who qualified for our leaderboards.
|2021 Passing Plus/Minus|
A Tale of Two Murrays
Murray led the league with a plus/minus of 27.4—so he completed 27.4 more passes than an average quarterback with the same types of throws would be expected to make. That is the lowest total to lead the league since we started keeping track of this in 2006, and remember, it's a counting stat—the 17th game should have led to larger numbers. Murray missed three games this season, so it's not a surprise that his total wasn't higher, but it is noteworthy that no one with a full 17-game season could top his total. Murray's 6.3% CPOE was better than 2020's leader, Deshaun Watson at 5.5%, but we're used to seeing the best passer in a given season up in the 7% or 8% range with Brees or Peyton Manning leading the way. Murray didn't win with a dominant season; he had a very good season in a year where no one destroyed the competition. That's not to take anything away from Murray, but it is interesting that we set an all-time low in a counting stat after expanding the schedule.
Murray's year was also very much a tale of two halves. In Weeks 1 to 8, Murray had a +22.1 plus/minus and a 9.4% CPOE; he was on pace not just to lead the league, but to set records as the Cardinals roared out to their 7-0 start. Then Murray hurt his ankle against the Packers and missed three weeks. When he came back, he was not the same player. In Weeks 13 to 18, Murray had a +5.3 plus/minus and a 2.6% CPOE—both respectable, but not league-leading by any stretch of the imagination.
Murray's drop in production lines up with more than his injury. The Cardinals started using fewer four-receiver sets and spread formations. Injuries along the offensive line continued to mount. And Kingsbury's offense, as it has throughout his time as a head coach, became more predictable and easier to cover over the second half of the season. But I do think you can blame a significant chunk of the dropoff on Murray. Being banged up limits the out-of-structure freewheeling that makes him one of the more exciting quarterbacks to watch, but he was also missing easy throws in December, the kind he had no problem making in September. He built up a large enough lead over those first seven weeks that no one had a chance to catch him, but he left the door wide open for the rest of the league to get close.
And while the injuries are obviously a factor here, it's fair to note that Murray saw his CPOE drop in the second half of 2020 as well. Over the last two seasons, Murray's +36.0 plus/minus in Weeks 1 to 9 leads the league, and his 7.4% CPOE trails only Russell Wilson. From Weeks 10 to 18, those stats drop to +10.7 and 2.4% respectively—still good, but more borderline Pro Bowler rather than superstar franchise passer. The second-half swoon of Kingsbury's teams is one of the stranger splits in the NFL at the moment. If they could find a way to keep their success going past Halloween, the Cardinals would be right up there with the top Super Bowl contenders.
Risers and Fallers
Right behind Murray was Joe Burrow, who is the passer with the greatest year-over-year improvement in CPOE, going from 0.2% to 5.4%. Burrow made significant strides in his second season, showing few ill-effects from the knee injury which ended his rookie season. It should also be noted that upgrading from A.J. Green to Ja'Marr Chase will help any quarterback. Burrow had a -9.2 plus/minus targeting Green in 2020; his CPOE for that year jumps from 0.2% to 3.2% if you ignore Green. Targeting Chase in 2021, Burrow had a +6.1 plus/minus, which is significantly better. But even if you throw Chase out of the equation entirely, Burrow would still have had a high CPOE—it actually would rise slightly to 5.5%. And you can argue that the only reason Burrow wasn't No. 1 is a lack of the kind of short screens that can quickly rack up points here—even if a throw would be completed 95% of the time, that's still a sweet +0.05 for your plus/minus for every completion. On throws beyond the line of scrimmage, Burrow actually outdoes Murray +24.3 to +22.1. Perhaps all Burrow needs to take the crown in 2022 is a Rondale Moore-type.
The other passer who saw a significant jump from 2020 was Carson Wentz, who leapt from -6.5% to -2.5%, meaning he moved from worst to fifth worst. I'd wager the vast majority of that improvement can be attributed to Frank Reich's offense rather than Wentz's talent at this point in time, and we'll see if that can hold up in Washington. The biggest fall from 2020 belongs to Baker Mayfield, who was progressively more battered and bruised over the course of the year. Ignoring him, the biggest fall would be Josh Allen going from 4.6% to 1.7%; 2020 was the kind of season that would be a career year for any passer, so some regression from Allen was perhaps to be expected.
Rookie Year Blues
We have to give significant credit to Mac Jones and, shockingly, Davis Mills for having very solid seasons in both plus/minus and CPOE. Rookies often struggle in their first year in the league, a byproduct of being drafted to bad teams and struggling to adapt to the speed of the NFL game. Jones is one of just three rookie quarterbacks since 2006 to hit double-digit plus/minus, joining 2012 Russell Wilson (+23.3) and 2016 Dak Prescott (+15.4); Mills ends up fourth just ahead of 2008 Matt Ryan (+8.5). You flip Mills and Jones around when looking at CPOE rather than plus/minus, as they finish fourth and fifth behind Wilson (6.4%), Prescott (3.5%) and, er, 2016 Cody Kessler (4.7%). So the fifth and sixth quarterbacks drafted last year were the best two. Of course. What else would you expect?
The other rookies had their struggles. Trey Lance didn't play enough to qualify, but he had mediocre numbers in San Francisco's offense (-0.2 and -0.3%, respectively). The other three first-year passers, however, all hit negative double-digits in plus/minus. Justin Fields was regular run-of-the-mill bad, but Trevor Lawrence and especially Zach Wilson were exploring new depths of poor play. Wilson's -36.0 plus/minus is the second-worst since 2006, and his -10.2% CPOE is fourth-worst. His processing has been excruciatingly slow, allowing defenders to close in and break up what should be easy checkdowns, leading to a lot of incomplete passes (or worse) on things that should be automatic. Wilson had 28 incompletions on passes with an average catch rate of at least 80%, most in the league.
Lawrence doesn't hit the bottom of any historic lists if you look at all passers, but he certainly does if you just look at rookies.
|Worst Rookie Passing Plus/minus, 2006-2021|
Ah, Blaine Gabbert, every poor quarterback's saving grace. Gabbert has saved so many people from being described as the worst ever in any capacity, he should be getting Thank You cards in perpetuity. Wilson gave Gabbert a run for his money in both plus/minus and CPOE, but Gabbert had the sort of generational badness you just can't beat.
There are quite a few names on this list who went on to be fine players, so a bad rookie season is far from a reason to panic. Lawrence's rookie numbers are very similar to Andrew Luck's, someone else who came into the league heralded as the best prospect in a decade. Of course, the 2012 Colts were slightly better in terms of talent and coaching than the 2021 Jaguars were, so Luck's overall performance ended up better, but that's what happens when you have Marvin Jones, Laviska Shenault, and Laquon Treadwell instead of Reggie Wayne, Donnie Avery, and T.Y. Hilton.
There's hope for the rookies, is what we're saying. More so for Lawrence than Wilson, perhaps, but one bad season does not a career make.
30 comments, Last at 18 Jul 2022, 4:26am
#1 by Aaron Brooks G… // Jul 15, 2022 - 10:22am
So the fifth and sixth quarterbacks drafted last year were the best two. Of course. What else would you expect?
Not totally unreasonable given that the top-3 QBs usually walk into raging dumpster fires with lines made of swiss cheese and a WR cadre manned solely by golems.
It's interesting that the list is dominated by midget QBs in a spread offense. What do the numbers look like if you don't ignore passes tipped at the line?
#28 by ImNewAroundThe… // Jul 16, 2022 - 11:13pm
Our fellow Aaron made a thread about it not too long ago. The correlation was the opposite actually but will people ever actually worry that a QB is too tall? Well as you see...no, it'll just be ignored because...doesn't fit intuitive narrative lol
Oh no shorter QBs compensate in other ways and...don't...let their passes get knocked down, which is...ba...well good of course! Still have to question if they can be successful/actually as good as the metric says though. Oh a tall QB gets them knocked down a lot, like Daniel Jones did in college? Pfffft, whatever, need context, oh and small sample size (only for them though)!!
I don't know why the football community still holds onto these pejorative myths (even the analytical side) when the league actually seems to be doing away/ok with it (Baker, Kyler #1 despite height/Goff, Burrow, Picket QB1 despite hand size. Hmmm what a wide variety of QBs, maybe there's more important things to look at).
#4 by BigRichie // Jul 15, 2022 - 11:51am
Why should +/- vary according to quality of receiver? No QB goes 'oh-wait-this-guy's-no-good-so-at-this-last-instant-I'll-now-put-less -effort-into-making-this-pass-accurate'.
Sorry, strikes me as a garbage stat. Good QBs are more accurate than bad ones, QBs who throw short are more accurate than QBs who throw long. Didn't need a brand new stat to tell me that.
Unless it points to +guys who the next year typically improve clearly more than the -ones, it's pointless.
#5 by Aaron Brooks G… // Jul 15, 2022 - 12:02pm
Why should +/- vary according to quality of receiver?
Why would it not?
It's comparing actual completion count against expected completion count. A roster of WRs who are better at catching the ball will improve that stat. A roster of WRs who are worse at catching the ball will depress that stat. Better receivers are better at executing catches than poor receivers.
#7 by BigRichie // Jul 15, 2022 - 1:00pm
Is that what it measures? My understanding was that the charters are saying 'so-and-so is on target XY% of the time according to our charts'. If not, what exactly is it that's being charted?
Sounds to me like you're maybe mixing up +/- with Zone %, for receivers.
#11 by Aaron Brooks G… // Jul 15, 2022 - 1:30pm
By "location" I believe they are referring to where on the field the pass originated and where on the field its destination was, and they tabulated expected completion percentage from those two inputs and compare the actual result on behalf of a given QB.
They are not documenting where on the receiver the ball arrived and the expected completion percentage of that.
That would make for an interesting analysis of receivers, though.
#13 by Bryan Knowles // Jul 15, 2022 - 1:36pm
This is where the confusion lies. When we talk about the location of quarterback passes, we're talking about direction and target depth, not whether a quarterback places it on the hands or not. This an analysis of play-by-play data as opposed to an analysis of charting stats.
So what we're measuring here is whether a quarterback is more or less likely to complete a pass at a given target depth, in a given direction in a given situation than their peers. That correlates a lot with accuracy, of course, but it also can account for things like decision making (did he find the open receiver, or try to force it into triple coverage?) and quality of the throw (did he lollipop a ball in, letting the defenders react, or did he put too much zip on it and bounced it off the receivers' fingers?).
#17 by KnotMe // Jul 15, 2022 - 4:24pm
Metrics are based on how often a pass is completed based on the pass distance, the distance required for a first down, and whether the ball was thrown to the left, middle, or right side of the field.
I think that is it. I imagine the first down distance matters, which is why it's included(it's not intuitively obvious). .
#20 by Vincent Verhei // Jul 15, 2022 - 8:24pm
"Direction" refers to three things:
- Depth of target (distance between the receiver and the line of scrimmage).
- Whether the pass was marked as left, middle, or right in the play-by-play.
- Yards to go for a first down. It's easy to complete a 5-yard pass on third-and-20. It's hard to complete a 5-yard pass on third-and-5.
#6 by Pat // Jul 15, 2022 - 12:51pm
No QB goes 'oh-wait-this-guy's-no-good-so-at-this-last-instant-I'll-now-put-less -effort-into-making-this-pass-accurate'.
Plenty of QBs go "oh wait, this guy's terrible, I'm not going to attempt this tight-window throw to him."
Plus you can go check out any former NFL QB's video breakdowns showing where apparently inaccurate passes are oftentimes the receiver's fault. Certain miscommunication plays are easy to tell, others aren't. Orlovsky specifically had breakdowns of Wentz's horrible, no good, very bad season showing plenty of cases where just minor things from the receiver lead to what look like very inaccurate passes.
#8 by BigRichie // Jul 15, 2022 - 1:06pm
They've always struck me as selective data-grabbing. Going in with a preconceived notion, then finding/seeing video examples thereof. So to speak, using the light+post for support rather than illumination.
#10 by Bryan Knowles // Jul 15, 2022 - 1:08pm
To add on to this, plus/minus is blame agnostic. Sometimes, an incomplete pass is because the quarterback overthrew his receiver by 23 feet, sometimes an incomplete pass hits the receiver in the breadbasket and he drops it. More frequently, the blame is on both players in some capacity - maybe not always 50/50, but an incomplete pass is usually at least a little bit the fault of both sides. Plus/minus is based strictly on the play-by-play data, rather than trying to make a ruling on every pass attempt in the league.
That being said, plus/minus and CPOE do align very well with charting results for quarterback accuracy. SIS' top five quarterbacks in on target% were Burrow, Murray, Carr, Cousins and Herbert -- 2nd, 1st, 5th, 6th and 19th in CPOE (to compare rate state to rate stat).
Herbert is the outlier there, which brings up questions about his receiving corps -- he had negative plus/minus targeting Mike Williams, Austin Ekeler and Jared Cook. Cook was second among tight ends in drop percentage, while both Williams and Ekeler were closer to the bottom than the top. So even when CPOE differs from charting numbers, it can raise interesting questions about why the numbers differ.
#12 by Aaron Brooks G… // Jul 15, 2022 - 1:31pm
More frequently, the blame is on both players in some capacity - maybe not always 50/50, but an incomplete pass is usually at least a little bit the fault of both sides.
It was your fault for thinking Agholor or Ebron weren't going to drop that pass.
#18 by mehllageman56 // Jul 15, 2022 - 6:48pm
Or Ty Johnson when your OC is telling you on the sideline to hit the open short passes. The thing is, the Jets running backs/receivers did that to Mike White as well (his two interceptions were deflected by his teammates).
So does this stat include receiver's drops or not?
#29 by takeleavebelieve // Jul 17, 2022 - 2:18pm
If you take that sentiment and reframe it as, “my first two reads were covered, and the #3 option who might be open downfield isn’t very good, so I’ll aim my throw to where it can’t be intercepted and maybe he’ll make a play on it”, I think it starts to seem a lot more reasonable.
#14 by Aaron Brooks G… // Jul 15, 2022 - 1:44pm
Was thinking about soccer statistics the other day -- XA for players, and XG for both players and teams. It occurred to me that the purest result for the analytics guys would be if games were decided not by score but by who won XG.
But it then occurred to me that if you did do this, player behavior would change because you changed the game statistic of significance. XG analysis only works because XGs are not actually the desired outcome -- real goals are. I would play differently if I wanted to optimize XGs instead.
It reminded me a bit of baseball -- baseball sought to optimize OPS, in part, I suspect, because it was easier to analyze and project true-outcome players. The perverse effect is it has depressed OPS in MLB -- we're now back to late 70's levels of offense.
The FO old-timers may have a take on this, but it would require narration. Did the introduction of sacks as an official statistic change how players played -- on both offense and defense? I could see that effect being either player-driven or front office-driven, but I'm curious if it existed or not. It was a bit before my time.
#23 by Bryan Knowles // Jul 15, 2022 - 9:00pm
The difference is basically all Jared Cook, who finished seventh-worst among tight ends at -4.4, as we'll cover next week. Suffice it to say, there's a reason Cook is no longer on Los Angeles' roster.