### Value Pricing the NFL QB

One of the things I have often done on my other website is work on something that I call “value price” analysis of positions. Essentially what I do is come up with different metrics, some existing and some my own, to create models in which we reassign market values throughout the league to various players. If you prod around the other site you’ll see models for pass rushers, quarterbacks, running backs, offensive linemen, and a few other items here and there. Today I wanted to touch on a new valuation for Quarterbacks.

Now before we start there are a few things to keep in mind. One is that these discussions are just to spur debate among readers to come up with new and different ways to merge the salary cap with analytics, two things that should go hand in hand. Two they are just based on one year of data whereas pure contracts should be based on a minimum of two if not three years. Third this will probably be long so feel free to skip to the charts and conclusion.

The Metric

In my prior work with QB’s I have looked at numbers from Football Outsiders, Pro Football Focus, NFL.com and my own Percent of Offense stat. I wanted to take a different approach this time and come up with a pure number from game data. The primary metric is derived from Pro Football Focus. PFF tracks for QBs success based on actual length of pass in the air. It is powerful data. For example a player who throws primarily down the field is only expected to complete around 35% of his passes while those who live short should complete nearly 73%. So what I have done is translate these numbers into expected vs actual performance, a metric I call incremental yards.

To illustrate the way this works we will use Aaron Rodgers of the Green Bay Packers. Rodgers threw 75 passes that did not travel past the line of scrimmage and they resulted in 0 interceptions and 413 yards. The average QB would have thrown for 434 yards and 0.48 picks. In this category we say Rodgers incremental effect on the game was -21 yards and a reduction of 0.48 picks. From 0-9 yards Rodgers was  268, 1908, and 3. The average QB would have been 1661 and 4.9, meaning Rodgers produced 247 incremental yards and reduced picks by 1.9. You should get the idea from here. We do the same for passes that travel 10-20 yards and 20+ and add them all together to determine that Rodgers produced 4,295 yards on a set of passes that an average QB last year would have passed for 3,906.7. So he added an additional 388.3 yards of passing offense to the team.

That handles the passing category but with so many QBs in the league now being used as runners we need to find out their benefit in that area as well. For that we turn to Yahoo where we calculate an average YPA, 4.28, for our QB field  and apply it to the attempts to calculate their run differentials. Robert Griffin ran for 815 yards on 120 attempts. The average QB would have run for 514 yards, giving RGIII a positive effect of 300.9 yards, which is outstanding.

You will also notice above that I mention interceptions which means I need to compensate for turnovers. We have the interception total now we just need to calculate fumbles lost. Again going back to Yahoo we can determine that our QB set lost 116 fumbles last season. To convert that to fumbles per play I use a “playmaker” base. The playmaker base is the amount of times a QB has the ball in his hands to make a play. This is dropbacks + run attempts – scrambles. Using that we say that a player should lose a fumble 0.59% of the time he has a playmaker opportunity. Football outsiders charts drive stats and has the average drive at 30.85 yards.  So every time our QB turns the ball over we are losing 30.85 yards. This rewards a player like Tom Brady who may not have the greatest passing numbers but never turns the ball over.  Brady threw 8 picks and lost 0 fumbles in 688 playmaker opportunities. He should have fumbled away 4.1 balls and thrown 16.3 picks. So his effect on the game is 383.1 yards saved just by turnovers alone.

The Results

Now there are two ways to look at this stat. One is the cumulative number which rewards players for 16 games and hurts those who were benched or injured. The second way is to break it down into yards per playmake opportunity. The following chart illustrates the each players total contribution, yard per chance contribution, and increase over the expectation for the player.

`NamePass YardsRun YardsTO YardsIncremental YardsYards Per PlayPercentGainedRobert Griffin III442.1300.9176.1919.01.69632.0%Cam Newton388.6196.952.6638.10.98518.2%Tom Brady293.0-66.5383.1609.50.88615.2%Aaron Rodgers388.327.6176.5592.40.90216.5%Peyton Manning366.0-92.5200.5473.90.75612.5%Colin Kaepernick219.3145.184.4448.81.50627.6%Drew Brees444.6-59.340.8426.20.59910.3%Russell Wilson284.586.322.9393.70.75614.0%Matt Ryan256.4-4.7116.1367.80.5439.2%Matt Schaub240.3-99.0141.9283.20.4798.5%Alex Smith243.6-0.825.7268.50.98419.0%Ben Roethlisberger123.6-19.4100.6204.80.4057.3%Tony Romo181.8-79.5-18.883.40.1172.0%Josh Freeman27.5-28.10.80.20.0000.0%Eli Manning-38.1-55.777.0-16.8-0.029-0.5%Kevin Kolb-109.631.427.6-50.6-0.225-4.3%Joe Flacco-62.5-115.1118.1-59.5-0.100-1.7%Sam Bradford-102.4-31.571.0-62.9-0.101-1.8%Nick Foles-85.1-5.1-3.9-94.1-0.318-5.9%Michael Vick-44.666.4-118.3-96.6-0.219-4.1%Ryan Tannehill-60.11.1-45.7-104.8-0.184-3.4%Jake Locker-89.4115.3-131.2-105.2-0.277-5.0%Carson Palmer-69.6-41.1-18.9-129.6-0.213-3.6%Brandon Weeden-75.8-4.7-76.6-157.1-0.275-5.1%Matt Hasselbeck-211.1-17.732.6-196.3-0.791-13.9%Chad Henne-98.5-17.4-85.2-201.2-0.567-10.3%Matthew Stafford-257.6-24.079.1-202.5-0.256-4.4%Jay Cutler-220.857.3-87.2-250.7-0.489-8.5%Philip Rivers6.4-75.7-182.5-251.7-0.417-7.8%Brady Quinn-201.0-15.4-57.3-273.7-1.155-22.2%Andy Dalton-102.7-81.4-90.7-274.7-0.442-8.0%Andrew Luck-272.2-10.6-26.7-309.5-0.424-7.3%Blaine Gabbert-301.9-21.1-8.1-331.0-1.041-18.7%Matt Cassel-131.229.3-309.1-411.0-1.269-23.3%John Skelton-264.2-12.1-146.0-422.3-1.920-34.6%Christian Ponder-360.3-4.1-79.5-443.9-0.772-14.3%Ryan Fitzpatrick-360.3-8.7-108.9-477.8-0.820-15.6%Mark Sanchez-387.0-66.3-332.9-786.1-1.544-27.2%`

RGIII graded out as far and away the most productive player in the NFL. If you threw another QB into that system and asked him to do what RGIII would be asked to do he would produce 919 less yards. Cam Newton was a distant second while Brady, Rodgers, and Manning round out the top 5. Interestingly enough Colin Kaepernick who did not play nearly the same amount of time as the other players ranked 6th.

The most average QB in the game was Josh Freeman while Eli Manning, generally regarded as elite because of his playoff success put up negative yards on the year. Most agreed that Manning had a down season and these stats lend credence to that. One of the more interesting numbers came from Joe Flacco who is now the highest paid player in the NFL. He produced -59.5 yards on the year, 17th on our list of 38 QBs.

And it would be no surprise to see the bottom of the barrel being Mark Sanchez of the Jets. Sanchez was essentially the anti-RGIII accounting for -786.1 yards compared to an average QB. That’s awful. Partially it is the Jets fault because no other team in the NFL would have allowed that to go on for 15 games. On a per play basis John Skelton was worse and got the hook.

When looking per play a few names stand out notably Alex Smith who ranked 4th in the NFL but was pulled in favor of Kaepernick, who was just outstanding.  The fact that two QBs performed so well in the 49ers system really points to how excellent of a job Jim Harbaugh did playing to the strengths of his players.  If Smith could do that for 16 games in Kansas City the Chiefs won’t regret those 2nd round picks one bit.

I also found Matt Staffords numbers to be eye opening. He puts up numbers but he also throws the ball a lot. Despite the weapons he isn’t having the success average players would in the same spots. Andrew Luck is also nowhere near the class of the other rookies at the moment.

Financial Modeling

The first thing I wanted to do was to set an overall market number for the league. The data was based on all QBs who took at least 25% of their teams snaps this year. 38 players qualified under that criteria. Of those 38 names 12 were new wage scale rookies and 1 (John Skelton) was a late round pick from 2010. Because their low APY’s skew the numbers I wanted them out of the equation. I also decided to pull out Josh Freeman and Joe Flacco (his old APY) since the mid round picks of the old CBA were paid well but not starter money like Matt Stafford or Matt Ryan received. Of the 23 that remained, 3 were signed to be backup QB’s (Brady Quinn, Chad Henne, and Matt Hasselbeck) so they were also pulled. I used the remaining names to calculate the average salary and consider that my baseline for an average player. The average was \$13,018,895.

To me that means around the NFL they should be willing, if everyone was a veteran QB, to have a minimum leaguewide set aside of \$416,579,040. Since we have an additional 6 players who qualified I wanted to add to the pool what I would consider average pay for a backup, which would be around \$3 million a year. That brings the pool total to \$434,579,040. Now I want to properly distribute that pool out to the players I evaluated.

In order to come up with a distribution I wanted to make two adjustments. The first was to make a pure scoring adjustment that would bring the lowest score up to 0, which would be the addition of 786.1 pts to everyones incremental score. I then wanted to ensure that the lowest point distribution would make up 0.345% of the pool, meaning their value would be equal to that of Brady Quinn’s \$1.5 million, the lowest that a team would pay for a viable backup. Here are the pay listings.

`NameCurrent SalaryAdjusted SalaryYPP Adjusted SalaryRobert Griffin III\$5,279,775\$23,052,175\$21,080,649Cam Newton\$5,506,375\$19,501,332\$17,230,371Tom Brady\$11,400,000\$19,140,435\$16,695,531Aaron Rodgers\$12,704,000\$18,924,221\$16,780,918Peyton Manning\$19,200,000\$17,426,172\$15,990,799Colin Kaepernick\$1,281,074\$17,108,298\$20,053,131Drew Brees\$20,000,000\$16,823,270\$15,143,819Russell Wilson\$749,194\$16,411,837\$15,989,408Matt Ryan\$11,250,000\$16,085,101\$14,839,650Matt Schaub\$15,500,000\$15,015,793\$14,492,525Alex Smith\$9,258,333\$14,829,941\$17,223,901Ben Roethlisberger\$14,664,417\$14,024,224\$14,088,794Tony Romo\$11,250,000\$12,490,721\$12,530,038Josh Freeman\$5,240,000\$11,438,551\$11,898,786Eli Manning\$16,250,000\$11,224,397\$11,739,324Kevin Kolb\$12,421,600\$10,797,183\$10,680,085Joe Flacco\$4,500,000\$10,684,039\$11,358,201Sam Bradford\$13,000,000\$10,641,868\$11,349,948Nick Foles\$692,130\$10,246,885\$10,175,403Michael Vick\$7,500,000\$10,215,710\$10,708,550Ryan Tannehill\$3,167,128\$10,111,929\$10,898,130Jake Locker\$3,146,501\$10,106,584\$10,397,810Carson Palmer\$10,750,000\$9,798,051\$10,744,544Brandon Weeden\$2,020,899\$9,451,280\$10,410,207Matt Hasselbeck\$6,666,667\$8,955,509\$7,610,876Chad Henne\$3,375,000\$8,893,740\$8,828,262Matthew Stafford\$12,250,000\$8,877,269\$10,510,958Jay Cutler\$14,668,750\$8,268,080\$9,250,877Philip Rivers\$15,300,000\$8,254,864\$9,640,206Brady Quinn\$1,500,000\$7,976,471\$5,641,811Andy Dalton\$1,303,550\$7,963,935\$9,501,247Andrew Luck\$5,527,000\$7,524,394\$9,600,999Blaine Gabbert\$3,000,412\$7,252,317\$6,259,385Matt Cassel\$9,669,800\$6,241,439\$5,026,871John Skelton\$530,413\$6,098,162\$1,500,000Christian Ponder\$2,539,675\$5,826,130\$7,716,437Ryan Fitzpatrick\$9,833,333\$5,396,734\$7,458,204Mark Sanchez\$13,491,667\$1,500,000\$3,532,387`

A few things immediately jump out. One is how good the young player pool in the NFL currently is. In part that is due to the running ability of the players and there is a question as to how long they can keep that up.  But if you can get that production from them early it is incredible value to have a player on a contract that pays him \$5 million or less a year when he would be earning closer to \$20 million as a free agent.

Joe Flacco really demonstrates the importance of the playoffs to a young QB. As a regular season QB he was a below average passer and run performer. He was safe with the football which is why he has a game manager label. But when you perform on the big stage teams can overvalue it. Now he has to prove everyone wrong and improve on what he did this year. If not he has a chance of being the most overpaid player in the NFL

Now on the downside numbers I would never expect teams to sign some of these players for these amounts. In theory I would think that any veteran player with negatives across the board would be relegated to backup duty or simply put out of the NFL. No team should pay Carson Palmer \$10 million simply because there is no upside to his game anymore. Maybe if you put him on a decent team the salary fits but on a squad like the Raiders it simply doesn’t. Cassel, Fitzpatrick, and Sanchez should all be toast.

Conclusion

This is just one of many ways that we will in the coming months evaluate a marketplace of players. Nothing is perfect and this certainly isn’t either.  The NFL is all about moving parts and systems. Joe Flacco was somewhat above average throwing the ball down the field. Eli Manning was terrific in that metric, 2nd best in the league to Drew Brees. A player like Matt Schaub who grades overall higher was awful in that regard. The Ravens and Giants would have to change up their entire philosophies to sign a player like Schaub. But if you have a flexible front office stats like these can give you an idea not just of gameplanning to strengths but also putting your salary cap dollars in line with what performance you are getting on the field.

As always feel free to email any comments or just post them below. Ive been ultra busy lately but always try to respond whenever I get the chance.

• I think there are so many variables that can’t be accounted for when looking at QB performance, a reliable “value pricing” model is going to be extremely tough to figure out. 6 stat categories don’t even come close to covering it all, and it is probably impossible to try to create a formula that accurately represents value.

That said, always great insight and well thought out and researched article.

• Thanks Pauly. I definitely agree. I think each team has different philosophies and different pieces that make it impossible to say player x fits in system y just because a number states something. Analysis like this also gives the QB all the credit (and blame) for yardage. Maybe Ray Rice just wasnt very good last year when used in screens or maybe the Ravens just call too many WR screens to slower wideouts. I also have no adjustments for O-line play(and its a significant factor).

I think the point for all of these that I have done on my other site and will do here is to just give some different ideas as to how to make some price points up on players. I could probably use a number of different tools to come up with completely different numbers, though I tend to think the selection of Skelton and Sanchez as the bottom two should be consistent everywhere- I lived through Sanchez last year and I have never seen anything like that before. As a 30 year Jets fans Ive seen some bad play but wow was that something last year. Just so bad on so many levels.

• Jason,

Great concept, but I think – crazy as it is to say this – it is based on too broad a statistic. Expected outcome per play will change based on defenses; a QB will have fewer expected yards and completions per play against the #1 pass defense than it will against the Patriots or Eagles. To make this useful you need to find some way to meld it with DVOA

• You know my first inclination was to adjust everyone for defenses played like I do my efficiency ratings until I realized how much of a pain it was to tabulate the defensive data without access to the raw data files (which I probably could have asked for). When Ive done this in the past I have found that most teams end up facing a relatively normalized schedule but there are certain divisions that throw that out of whack. With the data its actually simple to construct. I have the spreadsheets already built for it on a team basis but building the data seems pretty time consuming. 🙂 Ill look back into it though for V2 once free agency dies down and these contracts stop changing. Who knows maybe one day my play analyzer will see the light of day if I figure out an easy way to get data for all teams calculated with the click of a mouse. My last try at that froze up my computers 🙂