Correcting the NFL’s passer rating lie

  • last month
For more than a half-century, the NFL has used a complex, convoluted formula called passer rating to assess quarterback performance. But in an irrational quest to confine numbers within an unnatural, inorganic range, this results in an abundance of lies stemming from samples of individual games. This is a pursuit to shine a light on – and bring truth to – the deceit.

Category

🥇
Sports
Transcript
00:00Through the first 58 years of the Super Bowl era, there have been 27,686 regular and postseason
00:10team games. Those encapsulate a lot of performances, both good and bad, by a lot of quarterbacks.
00:19One of the traditional, fundamental ways to assess these performances is with the metric
00:24passer rating. Passer rating is a complex mathematical formula that incorporates four
00:30different variables, completion percentage, yards per attempt, touchdown percentage, and
00:35interception percentage. But there is one major flaw that to my knowledge doesn't exist
00:41in any of the other rate metrics in any of the three sports with which I'm familiar.
00:47It frequently lies. For each of the four variables, the formulas within the formula if you will,
01:00this metric decides it is simply unwilling to acknowledge numbers that fall below zero
01:05or above 2.375. This is important and cannot be overstated. That is an arbitrary, man-made
01:14barrier that was baselessly implemented and skews how we perceive passer rating figures.
01:20So based on this formula, and that quirky, artificial, random rule, if an NFL quarterback's
01:26performance in a given sample entails a completion percentage north of 77.5 or south of 30,
01:34it just pretends that those are the boundaries of the numeric counting system and rounds
01:38to those numbers since that's what they'd need to be to keep the result between zero
01:43and 2.375. If a quarterback completes every pass, the passer rating formula just decides
01:50that nope, I'm gonna pretend that actually 22.5% of his passes were incomplete. Likewise,
01:57if a quarterback tosses an incompletion on every pass, it just pretends he actually completed
02:0230% of his passes. For yards per attempt, no additional credit is given above 12.5 and
02:09the artificial floor below which the sliding scale inexplicably ends is 3. For touchdown
02:16percentage, since it's impossible to be in the negatives, the artificial and the actual
02:21floor are one in the same, but if a signal caller tops 11.875, sorry buddy, that's considered
02:28the highest number in existence here. And finally, for interception percentage, it's
02:34also impossible to be in the negatives, meaning the artificial and actual ceiling are one
02:39in the same, though on the other side of the coin, anyone who tosses picks on more
02:43than 9.5% of his passes is spared the ridicule of the true number being reflected.
02:49We'll dive deeper into the specifics shortly, but ultimately this has marred the passer
02:55rating listed for thousands of games. Artificial ceilings and floors are simply lies, and they're
03:01lies that have by and large been blindly accepted without question or pushback for
03:07decades. Now, it is true that across big samples of full seasons or careers comprised by several
03:15hundred or more passes, these bullshit boundaries don't really ever come into play. Over an
03:20extended period, no NFL quarterbacks complete passes at a rate below 30% or above 77.5%,
03:28and they don't average over 12.5 or fewer than 3 yards per attempt. They don't launch
03:34touchdowns on more than 11.875% of attempts, and with the exception of 1947's Sid Luckman
03:42and 1962's George Blanda, who predate this era anyway, they don't toss picks on more
03:47than 9.5% of passes. But on smaller, game-level samples, the bullshit boundaries come into
03:55play a ton. As an example, let's take the top game authored by my personal all-time
04:01favorite quarterback, Steve Young. On December 19, 1993, he completed 17 of 23 passes for
04:10354 yards, 4 touchdowns, and no picks. That amounts to a 73.9 completion percentage, 15.4
04:20yards per attempt, a 17.4 touchdown percentage, and obviously an interception percentage of
04:270. With the formula shortchanging Young in pretending he only averaged the upper boundary
04:32figures of 12.5 yards per attempt and an 11.875 TD percentage, that spits out an artificial
04:40rating of 155.3. But if the true numbers were utilized, 185.8, over 30 points better.
04:51In other words, the 155.3 is a lie. I don't like the lies. So I took a look at each and
05:00every game that each and every quarterback played in each and every season of the Super
05:06Bowl era, and I rectified it. Here are each of the era's 24,492 instances as of 2024 in which
05:17an NFL quarterback threw at least 20 passes in a regular or postseason game, a number
05:22representing nearly 90% of the overall team games, meaning only particularly small samples
05:29are omitted. They're plotted by season and by what their actual passer rating would be
05:35if we just let nature take its course and apply the accurate, truthful number for each of the
05:40four passing variables, even if they're below zero or above 2.375. If you squint really hard,
05:51you'll notice a few of them aren't quite readable. It's okay, we'll move forward anyway and do so
05:57with the benefit of mostly readable outliers. So just how prevalent are these phony barriers?
06:04Well, let's start with those outliers. With the artificial floor of zero for each of the four
06:10variables, that means a lowest possible numerator here of zero, and thus an overall passer rating
06:16artificial floor that exists for some reason of zero. That would erroneously lead one to believe
06:23that there was a 20-way tie at zero for the most inefficient passer rating games of the Super Bowl
06:30era. But instead we see that in reality, there were 178 games in which QBs finished a game with
06:37an actual passer rating in sub-zero land, and so this chart more properly separates and identifies
06:44the worst of the worst. On the other hand, with the artificial ceiling of 2.375 for each of the
06:52four variables, that means the highest possible numerator here is 9.5, which is then divided by
06:586 and multiplied by 100, and thus puts the overall passer rating artificial ceiling that exists for
07:05some reason at, and some of you dorks might already be familiar with this number, 158.3.
07:15So that's the math behind that, and that math would mistakenly lead one to believe there was
07:21a 38-way tie for the most efficient passer rating games of this era. But instead we see that in
07:28reality there were 201 instances in which QBs finished a game with an actual passer rating
07:34north of 158.3, and so this chart more properly separates and identifies the best of the best.
07:43Plenty more that weren't extreme outliers were nevertheless affected, for better or worse, by the
07:49artificial boundaries. Those 158.3s and 0s occur when all four metrics fall at or beyond the
07:57boundaries, but the passer ratings of more middle-of-the-road performances still become
08:01compromised if they happen to have 3, 2, or even 1 of the metrics fall beyond them.
08:08For instance, QBs have combined for 856 games in which they completed more than 77.5% of their passes.
08:17The 2.375 artificial ceiling from just that one element right there shortchanges nearly 900
08:24performances. None more so than 2019 Drew Brees' in Week 15 when he completed all but one of his
08:3230 passes. On the flip side, 137 player games came in under 30%, providing him with a bogus boost up
08:40to 30. None bigger than 1985 Dave Wilson, who opened the season by completing just 2 of 22
08:47passes before getting yanked less than three quarters into the campaign. More than 220 of
08:53these games featured a QB that averaged over 12.5 yards per pass, with the most harm in that
08:59aspect inflicted on a 1967 Johnny Unitas game in Atlanta where a full six yards per pass were
09:05chopped off for the purposes of passer rating for no reason whatsoever. Meanwhile, in a 1970 game in
09:12which in reality Billy Kilmer was swallowed up by the Purple People Eaters to the tune of not
09:17even matching his pass attempt total in passing yards, it just pretends that he had three times
09:23as many yards as attempts. Quarterbacks collectively had over 950 games throwing TDs
09:29on more than 11.875% of passes, with 1977 Bob Greasy not being thankful for drawing the ultimate
09:36short straw on Turkey Day when more than 26% of his 23 passes resulted in scores, only to be
09:43woefully under-recognized in the eyes of passer rating. And finally, in over 1500 games a quarterback
09:51threw an interception on more than 9.5% of his passes, with the trio that completed passes
09:56to the opposition on a quarter of their attempts getting the largest reprieve. Now while many of
10:01those aforementioned games within the four passing variables overlap, overall more than 3400, or
10:08about 1 in 7, of all these games are inaccurately portrayed by the NFL due to the nonsensical
10:15artificial bookends. With any simple percentage-based rate metric in any sport, it's impossible to be
10:21below zero or above 100, and that's always accurately reflected. If an NBA player hits
10:27all 10 of his shots in a game, we say his field goal percentage that game was 100. We don't
10:32artificially cap it at like 84.2 and call that or anything above it perfect because we don't like
10:39dealing with numbers that realistically could only stem from smaller samples. And there's
10:44absolutely no reason for the same concept to not apply to all rate metrics, even the more complex
10:51beyond a simple percentage. For example, historically the average big league baseball player has an OPS
10:57of roughly 725, but if one were to theoretically homer on every plate appearance of a given sample,
11:03his OPS therein would and should be characterized as 5000, regardless of whether or not that would
11:10look wonky. We don't randomly call an OPS of, say, 2800 to be perfect, with any performance
11:17beyond that rendered moot. Same on the other extreme. Though no one posts an OPS of zero
11:23across any sort of decent-sized sample, it's of course possible in a small sample. But there's
11:28not an artificial floor of, say, 300 to avoid wonkiness. People should have the option to
11:35embrace it. Besides, it's accepted that sample size would be inherently taken into account when
11:41dealing with rate metrics. That's why there are generally minimum qualifiers in connection with
11:47them. Ipso facto, the only ceiling that should exist within passer rating, the natural ceiling,
11:53and the number that should technically constitute perfection, is 831.25 in the theoretical scenario
12:00of every pass being a 99-yard touchdown. And the only floor that should exist, the natural floor,
12:07is negative 414.58 in the theoretical scenario of every pass being an interception. The artificial
12:15constraints of 0 and 2.375 within the individual variable formulas leading to the bigger picture
12:22artificial constraints of 0 and 158.3 are unequivocally irrational. Now, with that
12:30mathematical atrocity fixed, we're able to assess what we're left with. With the true figures revealed,
12:36they're teeming with takeaways and anecdotes. From the majesty of Lamar Jackson's game that
12:42sits atop the throne, having taken the elevator more than 65 points above so-called perfection,
12:49to the majesty of Dan Pastorini taking the submarine nearly 60 points below zero,
12:55and everything in between. With the emergence of clarity and the appropriate numbers unearthed,
13:01we find so many quarterbacks that have crafted so many performances far better or worse than
13:07what has been inaccurately conveyed for decades, and it's important to shine a light on them,
13:13which we'll continue to do in the episodes to come.
13:26Thank you all so much for watching and supporting Secret Base. If you haven't already, make sure to
13:31subscribe to our YouTube channel. If you want to get access to even more and see your name
13:35included alongside all these other distinguished folks, subscribe to Secret Base on Patreon.

Recommended