NFL Quarterback Performance After 12 Weeks

[Football Frontier RSS Feed]

There have been a few surprises in the NFL quarterback world this season. The biggest has been the sharp declines of both Peyton Manning and Andrew Luck. But both have probably played injured when they shouldn’t have. While they are … Continue reading →

More...

[PilotMan]

This thread, as innocuous as it is, has taken over my last couple days. In Jim's analysis he talks about taking his formula and adjusting it to account for the amount of a passes a QB takes in a game, weighting it as more valuable the more frequently a QB passes. Later in the post he talks about further analysis to include the amount of damage that a QB can do with his feet. So it got me thinking. How can I take the numbers that he has and what can I do with them to account for both variables? And what would the result look like.

First off, I suck at spreadsheets, so I had to do all these calculations and orderings by hand. It's taken me much longer than I should really admit. I took the basic score that Jim provided in his post then modified it to account for the number of times that the QB passes. So I started working on that with the idea, but I knew that I also wanted to create something that factored in the scrambling ability of the QB as well. So it made sense that a QB rush should be accounted for the same way that a QB pass is in the greater scheme.

For the group represented I found the mean of passes attempted as well as the mean for rushes attempted. Then I added the means of both categories together to create the total number of actual plays that the QB ends up being responsible for. Then for each player I took his actual combined number divided by the mean to create a multiplier for the score that Jim provided.

So QB's who run more, but pass less (Newton, Wilson) aren't punished as much as QB's who pass less and run less (Foles, Bridgewater). QB's who simply pass more still get held to the same standard, but it functionally treats a run as having the same opportunity as a pass. Looking at the numbers, the QB with the most plays was Andrew Luck with 47, compared to the group average of 37.354. In Jim's chart Luck is a league worst rating of 37. With my ajustment for his importance his score rises all the way to 46.55. Good enough to move him up to number 27.

Jim has 3 QB's at the top of the list tied with a score of 63, but with my numbers we see that Brady climbs out ahead with a number of 72.52, while Wilson and Palmer remain tied with 62.4 each. I'll have more on this tie later as these two separate. Tyrod Taylor, who Jim has in 5th with a 61, drops to 12th with his 55.52. He doesn't note it, but neither Cowboys QB qualifies. So Romo, nor Cassell are accounted for.

So Brady takes over the top spot, followed by Rivers, who moves up from 6th and then Brees who moves up from 8th. Hard to argue against that. Game manager Foles drops from 28th to 31st.

So here's my full list for the Jim's ratings, adjusted for the number of plays that the QB is responsible for.

Place Name Score Change in Rank
1. Brady, 72.52, NC
2. Rivers, 69.07, +4
3. Brees, 66.34, +5
4. Dalton, 63.07, NC
5. Roethlisberger, 62.64, +1
6. Palmer, 62.4, -5
6. Wilson, 62.4, -5
8. Ryan, 60.72, +3
9. Rogers, 57.83, +2
10. Smith, 56.46, -1
11. Stafford, 56.38, NC
12. Taylor, 55.52, -7
13. Carr, 54.93, -4
14. E Manning, 54.88, +6
15. Tannehill, 54.61, +4
16. Cousins, 53.49, -5
17. Flacco, 52.85, +8
18. Cutler, 52.04, -7
19. McCown, 51.16, +3
19. Bradford, 51.16, +3
21. Newton, 50.16, +3
22. Bortles, 49.5, +6
23. Mariota, 48.72, -6
24. Hoyer, 48.24, -8
25. Fitzpatrick, 48.03, +1
26. Winston, 46.85, -6
27. Luck, 46.55, +4
28. Bridgewater, 44.55, -9
29. Kaepernick, 41.23, -2
30. P Manning, 40.48, NC
31. Foles, 36.84, -3

In addition to Taylor, we see the rookies, Mariota and Winston take steep falls. The biggest drop in ranking belongs to Bridgewater whose tied with Foles for least plays, so his higher scores take a bigger hit versus the guys who are more important to their respective teams in terms of plays. I think we see the way that coaches control their younger or less talented players in this way, and it's reflected in the numbers.

The top of the list are the QB's who are generally considered the best in the business. Experiences QB's like Rivers and Brees are given a much bigger load to carry, therefore, their performance is considered more important. Palmer and Wilson are having great years, but they is responsible for 6-7 less plays per game than their counterparts.

Next I'm going to show you how I factored in the QB's running ability to further modify this number.

[PilotMan]

Trying to come up with a calculation that factored in the QB's running ability without giving it too much weight was challenging. I wanted to be able to give bonus points to the players who were really, really good, while not punishing anyone if they weren't big runners. I'd already taken into account the number of rushes that a QB takes in the first formula. Now I just needed to factor in the yardage.

I didn't do anything with scoring. While it's important it's somewhat arbitrary. Yardage gained is more of a direct representation of the scrambling of the QB. So how to go about it? After a lot of pondering and experimenting I settled on the following adjustments.

The first thing that I did was to calculate the average yardage gained by each QB per game. Then I found the average across the groups. Once I had that, I calculated a standard deviation for the group. What I found was that the average was 12 ypg rushing, with a standard deviation of 11.47. From here I put all the QB's into buckets based on the standard deviation.

QB's who were two SD's less were the baseline for the scores. Only 2 QB's fell into this category and averaged 0 yards per game, Drew Brees, and Peyton Manning. I didn't want to subtract anything from the adjusted scores that I had calculated, since running ability should be a bonus.

With each QB in a bucket the next plan was to come up with a number that would give enough of a boost to the guys who do scramble. Here's what I came up with.

I took the average of the adjusted rating from all the players. This number came out to be 51.98. From there I came up with a sort of award system based on which bucket each player was in. Players who were 1 SD higher than the baseline QB's would get a 3% bump from the average of everyone. Players 2 SD, 6%. A player 3 SD removed from baseline would get 12% and a player 4 SD higher would get an 18% bump. The only 2 players in the highest bucket were Newton and Wilson. By using the mean of the adjusted ratings as the basis for the percent increase for each player, it would ensure that the players in each bucket would get identical bonuses.

The reason that I didn't go with equal 3/6/9/12 bumps was that I felt like players who were really threats with their feet were that much more dangerous for opponents than those who were the most average. Players who were so far and away better deserved to be rewarded so the 3/6/12/18 made much more sense.

So in the end there isn't a whole lot of movement for the vast majority of the players, but the guy who do are the ones where the yardage they gain is important.

Clearly, with Wilson and Newton having the most to gain we see it reflected in their final scores. Wilson, who had been tied for 6th, jumped up to #2, while Newton who had been 21st climbed up the highest to 12th. Now, his importance is factored in as Jim had hoped instead of being way down in 24th.

No player was really hindered by a lack of rushing, there were only minor moves here and there. Here's how the final numbers look:

Rank Name Rating, Change from Original Ranking

1. Brady 74.08, NC
2. Wilson 71.76, -1
3. Rivers 70.63, +3
4. Brees 66.34, +4
5. Dalton 64.63, -1
6. Roethlisberger 64.2, NC
7. Rogers 64.07, +4
8. Palmer 63.96, -7
9. Smith 62.7, NC
10. Ryan 62.28, +1
11. Taylor 61.76, -6
12. Newton 59.52, +12
13. Stafford 57.96, -2
14. Carr 56.49, -5
15. E Manning 56.44, +5
16. Tannehill 56.17, +3
17. Cutler 55.16, -6
18. Cousins 55.05, -7
19. Flacco 54.41, +6
20. McCown 54.28, +2
21. Luck 52.79, +10
22. Bradford 52.72, NC
23. Bortles 52.62, +5
24. Mariota 51.84, -7
25. Fitzpatrick 51.15, +1
26. Bridgewater 50.79, -9
27. Winston 49.97, -7
28. Hoyer 49.8, -12
29. Kaepernick 47.47, -2
30. P Manning 40.48, NC
31. Foles 38.4, -3

Some things that stand out other than Newton's big jump:

Hoyer is the QB who takes the biggest hit in rankings. Ranked #16 in the original rankings his below average amount of plays per game and his below average scrambling ability push him down a total of 12 spots. One of the worst. Peyton's numbers were bad all around, his load is even below average for the group.

Andrew Luck is the worst of the original group, but his heavy load, coupled with his scrambling ability pushed him all the way up to 21. He stats still suck, but you can see that his season isn't as bad as all that.

Bridgewater, Winston and Mariota all fall down because of the limitations that their coaches put on them. They clearly aren't in a position to carry their respective teams and it shows. Take that against Bortles who has been rising on a lot of charts recently. He gaines 5 spots to 23, putting ahead of all three of the others. It showcases the teams increasing reliance on him as a leader.

Last, you've got Carson Palmer, who is having a great year, but it's just not worthy of the #1 spot that he help previously. His current ranking of 8th is much more representative of where he should be at.

I'm curious to see if anyone will even read this. Frankly, I'm glad it's done. It's been a little time consuming. Fun, but time consuming.

[albionmoonlight]

Originally Posted by PilotMan

I'm curious to see if anyone will even read this. Frankly, I'm glad it's done. It's been a little time consuming. Fun, but time consuming.

I've been following it. I think you've done a great job. And, of course, Jim for the original numbers and framework.

For years, I've been of the opinion that Brady has been over-rated. That it is more scheme and coaching and Gronk/Moss/etc. that have led to the Pats good numbers than the guy himself. (Really, ever since Belichick led Matt Cassell to 11 wins).

But, man, the guy just is unstoppable this season. What he is doing this year at age 83 (or whatever he is) is more impressive to me than the year when he threw jump balls to Moss and put up the videogame numbers.

Either he or Cam is the MVP. And it is close, but I'd have to go Brady.

[Solecismic]

This is the type of discussion that I want to work on. When I have a chance to work on the spreadsheets for a more extended time (mainly, I need to add sack/qb run data going back - I only have a couple of years of detailed data) I will start running correlations and try and extend my regression formulas.

I think (but I will check) the primary pieces I need to account for are:

Percentage of plays when the quarterback does not immediately hand-off the ball.

Quarterback rushing yardage.

I want to add these because I don't want to simply discount the success Newton is having despite traditionally poor quarterback numbers.

I don't like ESPN's leveraged situation approach. It assumes that somehow quarterbacks perform differently when plays are more important. Like baseball's "clutch" rating, I think all that does is discard important data and focus on tinier sample sizes.

Here's the complete metric score for this season. I cut off inclusion at 8 pass attempts in a game:

GP QBM Ave
Luke McCown 1 71 71.0
Austin Davis 1 71 71.0
Carson Palmer 11 696 63.3
Russell Wilson 11 694 63.1
Tom Brady 11 693 63.0
Andy Dalton 11 679 61.7
Tyrod Taylor 9 549 61.0
Ben Roethlisberger 7 420 60.0
Philip Rivers 11 656 59.6
Brandon Weeden 3 176 58.7
Drew Brees 10 585 58.5
Blaine Gabbert 3 175 58.3
Landry Jones 2 115 57.5
Alex Smith 11 632 57.5
Derek Carr 11 623 56.6
Aaron Rodgers 11 596 54.2
Jay Cutler 10 541 54.1
Kirk Cousins 11 595 54.1
Matthew Stafford 11 593 53.9
Matt Hasselbeck 4 215 53.8
Matt Ryan 11 590 53.6
Brian Hoyer 8 426 53.3
Teddy Bridgewater 11 572 52.0
Brock Osweiler 3 156 52.0
Marcus Mariota 9 466 51.8
Ryan Tannehill 11 562 51.1
Matt McGloin 1 51 51.0
Johnny Manziel 4 201 50.3
Jameis Winston 11 551 50.1
Eli Manning 11 549 49.9
Josh McCown 8 395 49.4
Tony Romo 4 196 49.0
Sam Bradford 9 440 48.9
Cam Newton 11 529 48.1
T.J. Yates 2 96 48.0
Joe Flacco 10 468 46.8
Matt Cassel 5 233 46.6
Ryan Fitzpatrick 10 456 45.6
Michael Vick 3 132 44.0
Zach Mettenberger 3 132 44.0
E.J. Manuel 2 88 44.0
Colin Kaepernick 8 349 43.6
Nick Foles 10 432 43.2
Blake Bortles 11 474 43.1
Peyton Manning 9 377 41.9
Ryan Mallett 5 191 38.2
Andrew Luck 7 260 37.1
Jimmy Clausen 2 64 32.0
Case Keenum 1 29 29.0
Dan Orlovsky 1 27 27.0

While tenths of a point are shown here, I generally report the findings as whole numbers because the metric itself is reported in round numbers. Over the course of an entire career, I think I should go down to tenths, because the metric can be calculated in tenths. But for now I'm staying with whole numbers. My intent is eventually (and this needs a lot of work) to put together a paper and hopefully present this work at an analytics convention.

[Julio Riddols]

I wonder how my QB rating system measures up to this.