≡ Menu

Guest Post: Bryan Frye and Win Contribution Rating

Bryan Frye is back with another fun guest post.  Bryan, as you may recall, owns and operates his own great site at http://www.thegridfe.com/, where he focuses on NFL stats and history.  You can view all of Bryan’s guest posts at Football Perspective at this link.

Oh, and Happy Thanksgiving to all the loyal Football Perspective readers!

Win Contribution Rating

It’s Thanksgiving. I don’t have a ton of time to write; you don’t have a ton of time to read. Let’s make this snappy.

A few months ago, I began using a rating that I feel better describes a quarterback’s contributions to helping his team win. I am terrible at coming up with names for stuff like that, but Football Guy Adam Harstad swooped in like a guardian angel and suggested the name “Win Contribution Rating.” I liked it, and I began using it without delay.

I used three metrics that correlate highly with future wins: Brian Burke’s EPA/P, Football Outsiders’ DVOA, and my Adjusted Yards per Play (AYP).1  The correlation coefficients with future wins (i.e., Year N+1 wins) for the individual metrics are .273 for EPA/P, .265 for DVOA, and .256 for AYP.2 When I ran those in a multiple regression, I got the following best fit equation (rounded):

Win% = .5 + EPA/P *.39 + DVOA * .13 + AYP * .008

Because the basis of this regression is win percentage, the equation spits out small decimals that I find aren’t relatable to most of the casual fans I know. To transform this into a number that resembles the NFL passer rating that people already know, I simply multiply by 140 to find the Win Contribution Rating.3

The highest score since 1999 belongs to Peyton Manning in his virtuoso 2004 performance. Let’s take a look at his rating:

EPA/P: .38
DVOA: 58.9%
AYP: 9.1
WCR = (.5 + .38 * .39 + .589 * .13 + 9.1 * .008) * 140 = 111.7

Here’s a look at the top 100 rated QB seasons from 1999-2013.4

12004Peyton ManningIND160.380.5899.11111.7
22007Tom BradyNE160.390.5418.12110.2
32011Aaron RodgersGNB150.40.4668.68110
42006Peyton ManningIND160.310.5137.57104.7
52013Peyton ManningDEN160.330.4327.75104.6
62009Philip RiversSD160.330.4177.48104
72005Peyton ManningIND160.330.4177103.4
82011Drew BreesNOR160.310.3837.97102.8
92013Nick FolesPHI130.30.3568.18102
102011Tom BradyNE160.310.3547.67102
112010Tom BradyNE160.270.4677.66101.8
122009Peyton ManningIND160.30.346.96100.4
132012Tom BradyNE160.270.3517.2699.3
142012Peyton ManningDEN160.270.3287.4499
152009Drew BreesNOR150.250.3677.2998.5
162007Peyton ManningIND160.260.3716.7398.5
172004Drew BreesSD150.280.3146.6798.5
181999Kurt WarnerSTL160.250.3697.2398.5
192004Daunte CulpepperMIN160.260.3516.998.3
202005Ben RoethlisbergerPIT120.240.3586.9197.4
212002Chad PenningtonNYJ150.220.4067.0397.3
222013Philip RiversSD160.230.3487.2997.1
232009Brett FavreMIN160.230.3457.2797
242009Tom BradyNE160.220.4046.796.9
252000Brian GrieseDEN100.220.3477.0796.2
262008Peyton ManningIND160.240.36.5395.9
272007David GarrardJAX120.220.3436.7795.8
282008Drew BreesNOR160.230.2866.8795.5
292008Philip RiversSD160.220.3037.0695.4
302000Kurt WarnerSTL110.230.286.8995.4
312000Peyton ManningIND160.20.3836.5795.3
322003Steve McNairTEN140.20.3636.5894.9
332000Daunte CulpepperMIN160.220.3016.5894.9
342010Aaron RodgersGNB150.220.2697.0694.8
352003Peyton ManningIND160.190.3716.8194.8
362010Ben RoethlisbergerPIT120.210.3116.6994.6
372006Drew BreesNOR160.230.2566.4694.4
382013Drew BreesNOR160.220.2696.7394.4
392010Philip RiversSD160.210.2796.9794.3
402005Tom BradyNE160.220.2866.2694.2
412002Trent GreenKC160.210.2716.9594.2
422001Kurt WarnerSTL160.190.3676.3394.1
432005Matt HasselbeckSEA160.20.3226.4494
442011Matt SchaubHOU100.210.2447.2194
452000Jeff GarciaSF160.190.3186.9894
462012Aaron RodgersGNB160.220.2346.8293.9
472007Tony RomoDAL160.230.2256.4893.9
482004Ben RoethlisbergerPIT140.210.3175.9493.9
492004Donovan McNabbPHI150.20.2757.0793.8
502009Matt SchaubHOU160.20.2936.7293.8
512009Aaron RodgersGNB160.230.1786.9293.6
522012Matt RyanATL160.240.1656.5993.5
532012Robert GriffinWAS150.230.1666.7993.2
542011Tony RomoDAL160.190.2687.0693.2
552005Carson PalmerCIN160.180.3346.4693.1
561999Peyton ManningIND160.180.346.3693.1
572004Trent GreenKC160.210.2646.193.1
582004Brett FavreGNB160.210.2456.3993.1
592003Trent GreenKC160.190.2816.4492.7
602012Drew BreesNOR160.210.1986.7192.6
612003Jake PlummerDEN110.190.2586.4492.3
622009Tony RomoDAL160.170.286.9292.1
632004Tom BradyNE160.180.3165.7492
642008Chad PenningtonMIA160.190.2116.6991.7
652012Russell WilsonSEA150.190.1976.4591.2
662008Matt RyanATL160.180.2535.9991.1
672002Rich GannonOAK160.180.2286.2791
682007Brett FavreGNB160.170.2456.2890.8
692012Colin KaepernickSF130.160.2586.4590.7
702010Peyton ManningIND160.190.196.0890.6
712009Ben RoethlisbergerPIT150.170.2326.3390.6
722010Michael VickPHI120.190.1416.7990.5
732000Rich GannonOAK160.180.2146.0690.5
742006Tony RomoDAL130.190.1895.990.4
751999Jeff GeorgeMIN120.160.265.8790
762007Ben RoethlisbergerPIT150.20.1275.9589.9
772006Donovan McNabbPHI100.150.1867.4289.9
782011Matt RyanATL160.170.1876.2989.7
792011Eli ManningNYG160.170.1626.5289.5
802009Eli ManningNYG160.180.1735.5989.2
812001Steve McNairTEN140.140.2615.9589.1
822010Josh FreemanTB160.170.1396.4489
832010Matt RyanATL160.170.185.5988.8
842005Jake PlummerDEN160.120.2716.2688.5
852013Russell WilsonSEA160.160.1566.1588.5
862011Philip RiversSD160.160.175.8388.4
872006Philip RiversSD160.160.1745.7688.4
882000Elvis GrbacKC150.140.2096.188.3
892011Matthew StaffordDET160.150.1496.4988.2
902008Jay CutlerDEN160.150.175.9287.9
912008Kurt WarnerARI160.150.165687.9
922005Trent GreenKC160.120.2595.8987.9
932012Matt SchaubHOU160.180.0755.9587.9
942013Colin KaepernickSF160.150.1665.9287.8
952002Jeff GarciaSF160.150.1545.887.5
962006Damon HuardKC100.120.2435.7787.4
972002Matt HasselbeckSEA130.130.2185.6887.4
982010Drew BreesNOR160.160.1335.4187.2
992003Matt HasselbeckSEA160.140.1735.6387.1
1002012Cam NewtonCAR160.180.026.187

Not many surprises here. The top rated seasons are seasons that most of us would agree are among the best of all time. Sixteen of the top twenty spots are occupied by surefire Hall of Famers, with two others coming from quarterback with an outside shot at a bronze bust. You also have Nick Foles’ 2013 and Daunte Culpepper’s 2004, which – anomalies or not – are undeniably great statistical seasons.5

Coming soon: A look at the quarterback performances of the current season. And, as always, please let me know your thoughts in the comments.

  1. Please note that the difference between the 45 yard penalty Chase uses and the 50 yard penalty I use for interceptions is based on this article by Brian Burke. I chose 50 as a compromise between the traditional and the new research. For fumbles, I used the standard 50 yard penalty and divided it in half to account for the randomness of recovery. []
  2. This includes all quarterbacks for which data is available from both FO and AFA, from 1999-2012. I did not include 2013 because I didn’t know year N + 1 wins; I’m not in the fortune telling business. []
  3. This may seem strange, but keep in mind that the NFL multiples by 16.67 to achieve its final passer rating. []
  4. Remember, the EPA/P and DVOA go back to 1999 and 1989, respectively. The dream is that we will one day be able to get these awesome metrics for older seasons; I imagine Marino’s 1984 might crack the top three. Unfortunately, I don’t think we’ll ever get a good look at Sid Luckman‘s 1943 or Otto Graham‘s 1953. []
  5. Note that when I say a quarterback had a great season, I mean a particular quarterback, behind a particular offensive line, with particular skill players and a particular defensive supporting cast, playing for a particular coach in a particular era had a great season. If you read this site regularly, you probably mean the same thing. []
  • jtr

    Since the stat is based on winning percentage, doesn’t it make more sense to just multiply by 16 and describe it in terms of expected wins (with average teammates assumed)? The passer rating scale is silly anyways–who wants numbers out of 167 or whatever? Interesting concept, though, to throw a couple of the best advanced stats together instead of just relying on one.

    • Thanks for the comment, and yes, it does make more sense to multiply by 16. That was actually my original formula (I use the word original loosely, given that the formula is a regression of other people’s work), but I ultimately decided to go with something that a casual fan would find familiar. I’m big on making things relatable, and even though the passer rating output is weird, people know it. I moved the multiplier to 160 before finally opting for 140. Either one is fairly arbitrary, but I found using 160 gave way too many scores over 100, which is something I am against.

  • Red

    Brian, I like what you’re doing here, but I believe you left out a very important component: Era adjustment. DVOA is adjusted for era, but EPA and AYP are not, so this metric will unfairly favor recent seasons where the rules have made passing easier. Comparing 2014 to 1999 is apples and oranges; it’s a different game with different rules. You know how Chase uses league average ANY/A in his VALUE metric? That kind of thing would work well here.

    • When I approached Brian Burke about using his metrics for this, he had the same concern about EPA/P, so you’re in good company with your criticism. He also pointed out that EPA/P doesn’t account for opponent like DVOA does. AYP doesn’t either, at least in its raw form. I share your concern, and I definitely don’t think comparing passing today to passing even as recently as 2006 is fair at all. Incorporating era adjustments is near the top of my list of offseason tasks. With work, school, and family, I don’t have much time to focus on working out kinks during the season.

      I use a similar concept to Chase’s RANY with adjusted yards per play which I call marginal adjusted yards. I do another one where I use one standard deviation below average as a baseline for replacement level.

      I’m interested in seeing how your work could be incorporated into these types of stats in order to build a more holistic rubric for quarterback grading.

      • Red

        I’d be happy to help you with establishing baselines for various metrics and creating “Over Average” versions of them. In fact, in would be awesome to build a database of these adjusted stats, because the NFL landscape constantly changes and raw numbers are basically meaningless. Chase can give you my email address so communication is easier. Happy Thanksgiving!

      • One more test for you, Bryan. Did you get a notification on this comment?