Passer Rating has largely become the industry standard in the NFL. It’s used in mainstream sports media, broadcasts and forum discussions alike. It gives fans and analysts a nice, clean number on which to judge a passer, rooted in quantifiable fact and not opinion. Aside from the subjectivity from stat-markers (e.g. whether to mark a gain for 8 yards or 9 yards, ambiguous fumble vs. interception cases, etc), there’s not much room to argue passer rating. Nobody says “in my opinion, Sam Bradford had a 95.6 instead of a 99.7″. That’s what makes it a useful tool. I personally cite passer rating all the time, and think it’s largely a reliable stat, albeit imperfect.
But many fans don’t know what passer rating actually is. They couldn’t tell you the formula, or what pieces it uses and in what way. They certainly couldn’t tell you where it came from. So let’s start this piece with a little rundown of passer rating, what it is, and why it works the way it does.
Up until 1971, the NFL had a tough time trying to determine a yearly passing leader. Was it the guy with the most yards? Best completion percentage? Touchdowns? Wins? It’s easier in a sport like baseball where everything that happens contributes to a run, and every run is one point. Football has 2 and 3 and 6 point increments, plus turnovers, and overall a much more complex system on which you have to measure progress. Then-commissioner Pete Rozelle commissioned statisticians, specifically a guy named Don Smith, to fix this problem. You can read the linked anecdote for more of the history behind passer rating, but at the end of it all, we get this:
That looks horrifying, but don’t fret- we’ll break it down. The first thing to do is note that it uses five raw numbers- completions, yards, touchdowns, interceptions and attempts. Not pictured are sacks, fumbles, rushing yards, throwaways or other classifications of QB-related play. We’ll get to that later.
According to the order of operations, we first have to solve the big complicated thing on top of the fraction. Starting in the parenthesis, we divide all four stats by attempts. This makes sense, as now you can measure QBs who played 3 games to QBs who played 16 in a nice, clean way. Saying “10 completions” doesn’t have a lot of meaning until you know if that’s 10 of 12 or 10 of 30.
After that, you have a problem of a bunch of different numbers on different scales. 0.07 TDs per attempt is actually pretty good, but 0.07 yards per attempt is outside reality. So how do you compare numbers on different scales? If you use their averages, you can scale them to comparable numbers. Passer rating scales these variables into values between 0 and 2.375, with 1 being average. Interceptions are scaled and then subtracted from 2.375 to reverse their impact and make them a negative.
Add them together, divide by 6, multiply by 100, and you’re golden. But why did we do that? Before we divide by 6, we have a number that would rank every QB exactly the same, so why the extra step? According to Smith, it’s to make the number easier to digest. A 100 is an excellent rating, and that’s a concept we’re familiar with. In school, anything below a 60 is a failure, 80 is decent, 90 is pretty good- it all bears out with our perceptions from formative days.
One thing that’s not shown is the limit on each value before they’re added up. This is more of a footnote meant to eliminate weird outliers. These limits are also why we get the odd maximum of 158.3. Don’t worry about it too much.
So we take four conventional things a quarterback can accomplish, divide them by attempts, do some math to them and add ’em up. So what’s the problem here?
The NFL has evolved quite a bit since 1973, when passer rating was officially adopted. An “average” mark across all categories in 1973 would be 1 in each category. 1+1+1+1 is 4, divided by 6, then multiplied by 100 for a value of 66.67. That’d be very bad in today’s NFL, where it wouldn’t even crack the top 30. Re-scaling each formula so that “average” is treated differently would make a lot of sense, though we’ll leave that project for another day. For now, we can just accept that the standard is higher for quarterbacks, and judge them in context of their peers.
What we can easily do to passer rating is update the definition of the variables. In this case, “yards” only references passing yards. But rushing yards are a legitimate part of the QB game now. In the 1970s, Fran Tarkenton was a pioneer and somewhat of a unicorn. Now, Russell Wilson, Aaron Rodgers, Cam Newton and Tyrod Taylor head a huge list of players who make their hay, in part, with rushing yards. So we’ll treat each rushing attempt as a passing attempt with the same result.
Attempts, currently, do not count all passing plays. A sack, generally a negative, is forgiven entirely by passer rating. So we’ll add sacks as passing attempts that went for whatever the net loss was. This probably underrates sacks, since a sack is more impactful than a gain of the same size, but including them is a step in the right direction. A fumble is also forgiven, unless it’s one of those ambiguous cases where it could have been a fumble or an interception. At that point, a statistician is making an ambiguous judgement call that shouldn’t affect passer rating at all. We’ll add fumbles to the interception total as though they were thrown picks. It may be better to use lost fumbles, but that would introduce the idea of fumble luck, and cause strange extremes with players like Ryan Fitzpatrick (9 fumbles, 1 lost) or Andrew Luck (6 fumbles, 5 lost).
So after all that, here’s how 2016’s quarterbacks (with 300 or more attempts) stacked up previously, where they stack up now, and what the difference in ranking is (e.g. Derek Carr drops 3 spots from 9th to 12th)
|Player||Passer Rating||“New” Rating||Rank Difference|
The biggest benefactors of this are, surprisingly, not particularly rush-oriented QBs. Remember, this still doesn’t care about volume. A QB can run for 350 extra yards, but if it takes 70 attempts to get there (5 YPA), that’s not a boon- nor should it be. Largely, QBs who don’t fumble or take an inordinate amount of sacks do well, like Eli Manning and Philip Rivers.
Big detractors are largely sack-magnets. A player like Russell Wilson, who has solid rushing numbers, takes way more of a hit because of his high sack total. A sack adds both an attempt without a completion and loses yards, so it hurts in two different categories. Carson Wentz and Andy Dalton suffer similar problems. Matt Ryan actually loses the 2nd most points of passer rating, but was so high above the rest of the field, he remains first.
The biggest detractor who leaps off the page is Sam Bradford. His astronomical sack total, complete lack of rushing game and high fumble total makes his passer rating plummet. He drops from 6th to 22nd, a fall of 16. For fun (and since this is a Vikings blog), I calculated Teddy Bridgewater’s 2016 totals. He falls from 88.7 to a 67.2, which is another huge drop due to sacks. He remains below Bradford, though the gap does narrow.
There’s a lot more work to do with passer rating in weighting all these factors. We improved the variables, but the math itself is still based on outdated relics. The league has changed a lot since the post-merger frenzy that birthed passer rating. Stats like ANY/A, QBR and DVOA are still better, though not as popular. We can also find a way to incorporate first downs, like Brian Frye’s TAY/P so masterfully does (seriously go look at that article). But adding those factors would make passer rating cease to be passer rating, and lose that marketability. The least we can do is update it to include all the ways that quarterbacks quantifiably contribute to a game.
Thanks for reading!