Introducing the PFF Passer Rating

| August 15, 2011

A few months back, I took a look at a number of statistics and problems in the way that they are used in a series called Stat Sheet Misconceptions. A number of these dealt with common statistics for quarterbacks, leading to a discussion about the current passer rating and the problems I have with it. At that time, I didn’t offer a specific alternative, but suggested that pretty much anything would be an improvement.
.
Fast forward a few months to now, and quarterback ratings have come into the spotlight again with ESPN’s new Total Quarterback Rating being released. It definitely is an upgrade over the past system, but because ESPN hasn’t revealed the exact formula for it, it’s been difficult for it to fully catch on. In the time since the our Stat Sheet Misconception pieces ran, we’ve built another alternative to propose one more step in the right direction.
.

Why Another Rating?

You might be wondering … why bother offering a new passer rating when we already have our PFF passing grades and ESPN has put out their new rating? Both are dramatic improvements in evaluating passers and look nothing like the standard that has been in place for way too long. Also, while only ESPN can produce the numbers for their rating, ours is dependent on information found only here at PFF. Conversely, with this effort, we’re aiming to produce a formula that anyone could use – given the right inputs – as an alternative to the standard.
.
.

The Old Formula

While the NFL’s passer rating is a set of complex formulas with upper and lower bounds, in most cases you can ignore them, and the formula simplifies to this.
.


.

The goal in creating the new formula was to maintain a similar structure but introduce values that make more sense. If you interested in why we’re changing parts of the rating, the Stat Sheet Misconception article on passer rating will help. Here we’ll stick with focusing on the changes being made .
.
.

Changing the Inputs

The first Stat Sheet Misconceptions article dealt with completion percentage, and offered a better alternative we called “Accuracy Percentage”. The first step in fixing the Passer Rating formula was to upgrade the Completion Percentage component to PFF’s Accuracy Percentage. With this move, dropped passes, throw aways, and spikes are now accounted for.
.
The second alteration was replacing Yards with “Yards in Air” to reflect the distance that completed passes travel from passer to target. While a quarterback has some effect on yards after the catch, this is difficult to quantify, so we’ve opted to boil it down to the point at which they surely have control.
.
.

Changing the Constants

We eliminated the “+2.083”, because it was simply not needed. To fix the issue of overvaluing touchdowns and interceptions in the old formula, the constants for each were lowered, to 20x for touchdowns and 45x for interceptions. Those values are more consistent with what others have found, and they worked well in this formula too. The last constant-related change was moving the multiplier from 4.16667 to 4.6667; a slight difference, but it adjusts the numbers to be more in line with the look of the old rating.
.
.

The New Formula

Applying it all, here is what we end up with:
.


.
.

Results

Below are 2010’s Top 10 quarterbacks as judged by the new PFF Passer Rating. Also listed are their marks in the old passer rating, and their ranks in each.
.

Pro Football Focus Passer Rating

Rank
Player
Team
PFF QB Rating
Original QB Rating
Original Rank
1Aaron RodgersGB98.33103.192
2Tom BradyNE97.95109.151
3Vince YoungTEN96.9398.644
4Michael VickPHI92.4998.395
5Philip RiversSD92.13101.823
6Peyton ManningIND91.3292.4110
7Drew BreesNO90.7091.2412
8Josh FreemanTB90.2295.866
9Joe FlaccoBLT89.5393.738
10Matt SchaubHST89.3192.0311

.

Outside of the Top 10, Colt McCoy’s rank skyrockets from 33rd to 17th, largely due to the facts that 11 of his 222 passes were thrown away and another 18 were dropped. On the flip side, we see both Jon Kitna and Jay Cutler fall eight ranks. Kitna had a relatively very low percentage of his passing yards before the catch, just 42% of his yards were in the air. The new rating helped other QB’s more than it did Cutler in part because he had so few passes dropped.
.
.

Closing Thoughts

While this new formula certainly isn’t perfect, it addresses a lot of the problems from the old rating system and it’s a bit easier to calculate than the old rating with no maximums or minimums for each value. To note: it’s capable of producing a max rating of 648.67, in which case every attempt would be a 99 yard touchdown with all the yards being in the air … so really there is no maximum.
.
If you asked me what the best method for rating passers is, I would still defer to the PFF pass grades. This exercise has not been about revolutionizing the passer rating but rather offering an improved alternative that deals with the problems from the past.
.
.

Follow Nathan on Twitter: @PFF_NateJahnke and check out our main Twitter feed too: @ProFootbalFocus
.
.

.
.
.

  • snowman88

    Love it

  • speirs

    Why not just subtract the drops in the denominator?

    • Nathan Jahnke

      The reason we have thrown away passes and spikes in the denominator is because they are plays that have no intention of being completed. Our goal in the passer rating is to only evaluate quarterbacks on passes that are intended to be complete.

      For drops, those are intended to be complete. In our opinion, the quarterback has done what he needs to do in order to have a positive play, but the the receiver is the one who messed things up. If we were to put drops in the denominator, then those plays wouldn’t be included in the rating. I believe instead of being irrelevant, that quarterbacks should get benefited because they did what they were supposed to do on this play.

      Something I should have mentioned in the article but forgot, not only are drops included, but yards in the air on dropped passes are a part of YIA. Basically we’re treating dropped passes as completions when it comes to quarterbacks.

  • PaulK

    A pretty good interception is a ball thrown 40 yards down the field on third down, when your offense needs points. A bad interception is an out route in the red zone which gets picked off for 7 points the other way. There’s a difference.

    Your system shorts Tom Brady, who leads his receivers well, so that they pick up extra yards after the catch.

    • Rick Drummond

      Hey Paul,

      Including yards after catch would credit him with work he’s not done, so the best option is likely somewhere in between YIA and all yards but finding that fair point is tough, especially when aiming for a formula that can be applied to all (50% of yards after the catch belong to the QB? 20%? 5%?). So, like mentioned, it was decided to limit it to YIA that it could be more clearly linked to the efforts of the QBs.

      Yes, Brady probably does get shorted by the decision to limit the yards considered here to those in the air, but the same could be said for many others and Brady, relatively speaking, isn’t hurt all that badly. With eight QB’s enjoying more YAC/comp last season (and “suffering” here more than he), it would appear he’s not the only one that effectively leads receivers to open space – or depends on his receivers to create after the catch.

  • baronzbimg

    I had already dropped a comment on the stat sheet misconceptions article but I’ll go at it again here. I love everything on this site, except for the irrational bashing of the best stat in football.
    While passer rating isn’t ideal to rate a player’ real level of play (defenses can drop 3 interceptions in a game on awful throws by the QB), it’s nonetheless an amazing stat because it’s related to the only thing that should matter in football, winning. Teams that won the passer rating battle last year were 203-53(!!), so in other words teams that posted the higher passer rating won 79,3 % of the time. Isn’t football about winning ?
    Who is the highest rated player in playoff history ? Bart Starr, the only 5 time NFL champion. Who is the highest rated Super Bowl passer ? Joe Montana.

    Labelling the new ESPN rating as “a huge improvement” is almost laughable. That rating had Dan Orlovsky above Ben Roethlisberger in 2009.

    Please read those 2 articles on coldhardfootballfacts.com, the apostles of passer rating:
    http://sportsillustrated.cnn.com/2011/writers/kerry_byrne/08/03/defending-qb-rating/index.html
    http://www.coldhardfootballfacts.com/Articles/46_3865_QBR%3A_ESPN%27s_Deeply_Flawed_Made-For-TV_Stat.html

    The best way to rate players is clearly what you do here on profootballfocus, rating every single down. But dissing passer rating has absolutely no sense, it’s the best stat in all football, and it should be used much more than the volume stats every american writer seems to enjoy. It’s all about efficiency, and efficiency leads to … victory.

    • Nathan Jahnke

      I think the biggest disagreement I have with you and Kerry Byrne’s article is that you’re looking for statistics to correlate well to winning, where we are trying to find the best methods to evaluate players, and I don’t believe those two are the same thing.

      Something like touchdown passes, or comparing touchdown passes from one team to another will clearly have a large correlation to winning. Thats because this directly leads to points, so it makes sense that teams with more points when more games. The passer rating gives a huge bonus for touchdowns. However if I were to watch every 35 yard completion a quarterback had in the past few years, I’m guessing I would find the majority of those to be great throws and should be positive plays for the quarterback. On the other hand, one yard touchdown passes are good plays, but I would argue in most cases the 35 yard passes are better plays a quarterback could make.

      The problem is, the quarterback rating favors the 1 yard touchdown pass over the 35 yard completion. If it was the other way around, it wouldn’t correlate to winning as well, but it would correlate better to who was actually a better quarterback. If I wanted to just have a quarterback statistic that correlated well to winning, I would include the number of kneel downs and have that be a huge bonus, and then I would have a number even more correlated to winning then the current passer rating. That simply isn’t our goal, not just here but in all of our player ratings.

      With the article in response to ESPN’s, they first argument is the historical thing. While I agree it would be better to have a system that works in any era, this isn’t something that ESPN or I tried to do. The numbers I use here only go back to when we have data, which is 2008, and ESPN’s also only goes back to 2008. Clearly not what we’re trying to fix with the attempts to improve the rating. The other thing brought up is the maximum score being an odd number, which I don’t have a problem with, and agree ESPN made that part a much bigger deal than need be. While I think how ESPN presented it was meant for the general fan, their methods behind how they came up with their numbers are a lot more complex then they showed on the screen.

      In your last point, I agree the volume, or cumulative stats like yards should be used less in favor of efficiency statistics. I just don’t think that the current quarterback rating is about efficiency… it’s more about big plays and high completion percentage.

      Thank you for being respectful in your criticism. I believe it’s a difference in philosophy on statistics.

  • http://www.profootballfocus.com Mike Clay

    After baronzbimg’s post, I was going to respond, but Nate basically said what I would’ve said right off the bat:

    I think the biggest disagreement I have with you and Kerry Byrne’s article is that you’re looking for statistics to correlate well to winning, where we are trying to find the best methods to evaluate players, and I don’t believe those two are the same thing.”

    That’s exactly right. QB Rating predicts Wins and Losses well because it’s not just a QB stat…it’s partially a team stat. Tom Brady could throw 10 passes directly to David Harris, but if he drops them all and Brady ends up with a 109 QB Rating, the Patriots will likely win. On the flip side, Mark Sanchez could complete only half his passes because his receivers dropped 8 balls, which results in a 62 QB Rating. In this very basic example, we see why the Patriots won the game, and also why QB Rating is a partially a team stat.

    An adjusted QB Rating like we have here at PFF removes some of the team impact and gives us a better metric. Perfect? Of course not. There never will be a perfect metric for this. But it’s definitely better.

    In fact, I had thought about expanding on this rating by normalizing based on both opponent and average length of pass attempt. That’s the next step.

  • http://www.profootballfocus.com Mike Clay

    Nate, Shouldn’t “hit while throwing’ and ‘batted balls’ also be removed from the denominator? It’s a tiny adjustment, but as far as I know, the PFF Aimed Passes = Pass Attempts – Hits – Throw Aways – Spikes – Batted Balls

    • Nathan Jahnke

      While those could definitely be considered, I could also see an argument for leaving them out. For both passes thrown away and spikes, when he began the process of throwing the ball, he intended it to be an incomplete pass. For batted passes, I could see the argument that a quarterback should’ve thrown the ball at a higher angle, or should be aware that a defender could be in position to bat a pass. For hit while thrown, then the quarterback should’ve been aware that he was being pressured and thrown the ball earlier.

      I think if I added those to one it would be for accuracy percentage, since I can for sure see an argument for it there. We’re looking for how good the quarterback is at hitting his receiver, and since those passes never get to the receiver, his accuracy shouldn’t be penalized for it. I’ll think about that one for a bit and if anyone else has any feedback on that I’m all ears.

      • http://www.profootballfocus.com Mike Clay

        Okay, I can buy that. Thanks.

  • baronzbimg

    Hi guys, I do respect your job a ton and maybe removing spikes etc … will actually improve passer rating. Passer rating will never be as precise as your ratings when it comes to judging a QB actual performance, but I really think a high passer rating (in the long term) absolutely translates good QB play.

    It seems we all agree in the end …