Week 3 Expected Points Added: Quarterbacks

This is the non-fantasy version1 calculation for the actual value – in terms of points the number of expected points added (EPA) – by quarterbacks so far in 2017.

I’ve made some adjustments to the formula in an attempt to better capture the value we can attribute to quarterbacks specifically, rather than assign  them the entirety of the value added from their teams’ passing plays.

ESPN’s Total Quarterback Rating (QBR) gives a broad outline of how they adjust and portion the EPA given to quarterbacks. I’m incorporating a lot of their principles: giving more credit for air yards versus yard after catch, dividing the negative points for sacks and inceptions between quarterback and linemen and giving the quarterback points for expected points gained on defensive pass interference penalties as if they were completed passes.

I also incorporated some of the methodology from Pro Football Reference’s Approximate Value calculation in using NFL salary cap allocations to determine how to allocate EPA among different positions.

By my calculation quarterbacks, receivers, offensive linemen and tight ends account for roughly 30%, 30%, 30% and 10%, respectively, of the salary cap spent on offensive passing in 2016.2 I used those rough numbers and intuition to come up with the following allocations of various types of EPA to the quarterback:

  • Air yards EPA: 40%
  • Yards after catch EPA: 20%
  • Interception EPA: 40%
  • Sack EPA: 40%
  • Rush EPA: 80%3

The air yards and yards after catch EPA are incrementally higher and lower than the baseline 30% based on the fact that the yards gained through the air are more about the quarterback, while the yards gained through YAC are more attributable to the receiver.

I also incrementally raised the interception and sack EPA attributable to the quarterback because the receiver is less of an influence for interceptions, and sack as much more a reflection on the quarterback than most think. There are many examples of quarterbacks who maintain an extremely low sack percentage year after year,4 no matter how who compromises their offensive lines.

I went fairly high for rush EPA only because I believe that the success of scrambling and quarterback runs is largely based on the vast differences in rushing ability between quarterbacks. Plus, quarterbacks runs are largely zone read based, so the quarterback is also making the choice to run, so the quality of the offensive line isn’t as important as on a pre-determined running back rushing attempt.

What my EPA calculation does not include is QBR’s “clutch” weighting. First, I don’t really like the idea of given more value to a portion of an already small sample. Second, I’m not really equipped to accurate estimate what the adjustments should be based on score differential and time remaining. While I agree that some players5 are garbage time heros, I don’t think that defenses simply give up yardage unless the games is entirely out of control.

Thinking about these numbers in a larger context, research from Max Mulitz has shown that you need roughly 36 in point differential for each additional win above a baseline of eight.6 My expected points calculation estimates that Tom Brady, the top quarterback so far in 2017, has added 14.4 expected points in three games. If we extrapolate that rate out over a whole season, Brady would be responsible for around 77 points added, or a little more than two wins. By this logic, adding Brady to an otherwise average, 8-8 team would bump their expected record to 10-6. That might not seems like a lot, but two additional wins from only one of 53 active players is significant.

With all this said, here are the updated quarterback EPA calculations for 2017 powered by the nflscrapR package. The total EPA numbers are upfront, and the components are broken out thereafter. Again, the table is sortable/searchable to your desire. The table is initially sorted by total EPA.

PlayerTeamEPAPlaysEPA/PPass EPAInt EPAPI EPASack EPARun EPA
T.BradyNE14.41270.1120.101.6-8.20.9
J.GoffLA8.5890.111.9-1.71-1.6-1.1
C.KeenumMIN6.6760.09600.7-1.31.2
C.WentzPHI5.41400.048.8-4.51.9-4.84
J.BrissettIND5.1820.064.8-2.20.8-2.64.3
B.BortlesJAX4.91000.053.9-2.42.1-2.43.7
M.RyanATL4.71030.0514.3-7.80.9-4.90
M.StaffordDET4.41220.047.9-4.51.3-4.50
J.WinstonTB4.4770.0610-5.50.8-2.40
A.SmithKC4.11050.0410.101-8.21.2
D.BreesNO3.91130.037.300-3.40
M.MariotaTEN3.51150.032.5-1.81.2-1.32.9
D.WatsonHOU2.91100.033.4-53.3-6.37.5
T.TaylorBUF2.71070.037-2.10.6-5.12.3
B.RoethlisbergerPIT2.11210.023.5-1.83.6-4.80
J.McCownNYJ21020.026.5-1.90.4-6.43.4
D.CarrOAK21020.024.2-2.42.8-3.90
P.RiversLAC1.91220.028-6.53-1.90
T.SiemianDEN11220.019.3-8.21.6-5.40
D.PrescottDAL0.111903.8-6.20.4-24.1
R.WilsonSEA-1141-0.01-0.400.3-5.24.3
A.RodgersGB-2.2162-0.0111-7.72.5-9.81.8
C.PalmerARI-2.5153-0.027.3-8.72.5-5.11.5
J.CutlerMIA-2.784-0.030.4-1.90-1.20
K.CousinsWAS-2.8113-0.027-1.61.2-9.2-0.2
D.KizerCLE-3.8138-0.035.2-11.92.6-5.96.2
C.NewtonCAR-4.2104-0.044.8-7.60.6-6.54.5
E.ManningNYG-5.1126-0.046.6-7.60.9-50
J.FlaccoBAL-9.375-0.12-0.3-5.40.5-4.10
B.HoyerSF-11.8113-0.12.7-7.60.4-8.31
M.GlennonCHI-13.4117-0.113.2-7.10-7.5-2
A.DaltonCIN-14.3109-0.131.1-8.30.6-8.60.9
  1. Aka “real” football
  2. This assume a 60/40 split between passing and running
  3. Includes scrambling
  4. See Peyton Manning and Dan Marino
  5. See Blake Bortles circa 2015
  6. i.e. An average 8-8 team

Leave a Reply