Notes on the quarter-back position
There are 32 quarterback jobs in the NFL, and 28 men who have shown the ability to play the position at all over the past two years. Here is the list, with combined 2015 and 2016 statistics (sorted by harmonic mean of TD/INT, Rate, and IAYPA ranks). I include Dak Prescott, taking his 2016 season as down payment on a nice future career:
There are four teams that will have to play someone not on this list - a rookie, or Gino Smith or someone. Some of the teams at the bottom are already moving on to their next option:
- Brock has been traded by the Texans to the Browns, who will either trade him again, or cut him. David Carr: "Bill [O'Brien]'s system works, and I don't think he's coaching it poorly. But in the times we live, it's going to be difficult for him to have enough patience to stick with one guy. And it's not just the quarterback. It's the combination of quarterback and receiver. Julian Edelman practically lives with (Tom Brady) during the offseason. They go through game situations three times a week. How many times did Brock do that? Probably never."
- Ryan Fitzpatrick will not re-sign with the Jets.
- Flacco is not only one of the three least productive starters of the past two seasons, he will, if Tony Romo is cut or traded, become the player with the highest salary cap number in the NFL. No sign his job is in jeopardy...Ozzie Newsome is now reportedly looking to sign some receivers to see if that would help.
- Blake Bortles - New coach Coughlin is not a fan.
That gives us at least eight teams - a quarter of the League, where there is not clear incumbent, or the incumbent has been awful for two years.
In the middle, Matthew Stafford, Alex Smith, Marcus Mariota are the median NFL quarterback, and therefore extremely valuable.
Not far south you can see Colin Kaepernick, who had the fewest starts of anyone but Prescott, but has actually performed well. INT/Attempt of 1.6% is well below median of 2.1%. For all the criticism, he is better than what about 1/4 of the League has.
A bit further down we find Cutler. He is the median quarterback on IAYPA, but his TD/INT ratio shows why Bears fans became exasperated. Andrew Luck threw 46 touchdowns in 22 starts over the past two years, Cutler only 25 in 20. I still think he could play somewhere, but his numbers are fully consistent with the criticism - he moves the chains but not the scoreboard. But again, do the Bears really think Glennon is going to be better?
The view from the summit:
- The Legend of Tom continues. The best quarterback in football over the past two years, by a lot. The typical competent NFL quarterback throws about twice as many TDs as interceptions (2.3 median for this group). Brady's ratio the past two years has been 7.1...
- Dak Prescott had as great a rookie seasons as Dallas could have hoped for.
- Russell Wilson is not just good, but great, much better than people seem to realize. This article says the offensive line needs repairing, but, like Rodgers, Wilson has shown he can perform at the highest level even with a weak supporting cast.
- Rodgers' IAYPA has slipped, but he is, as Simmons says, the moral equivalent of Omar from the Wire: "Rodgers is Omar. He’s a one-man gang. You’re always afraid of him, you can’t ever count him out and you never know when he’s coming. And those three Hail Marys (THREE!!!!) were football’s equivalent of Omar escaping Marlo’s crew with the five-story balcony jump."
- Drew Brees - just two years younger than Brady, and has more yards than anyone the past two seasons, with elite-level efficiency. No sign of decay yet.
- Kirk Cousins, who is
demanding a tradeon a one-year contract, is now one of the best quarterbacks in the NFL.
I've never seen it all so wide open. One NFL GM commented on the Bears' signing of Glennon: "I’d take Glennon over Osweiler in a heartbeat. He won’t do it himself, but if he has people around him, he’s got a chance. And that’s three-quarters of the league at quarterback."
5 Comments:
WDYJ (why don't you just) factor in touchdowns in your YPA calculations, then you don't have to take the harmonic mean (whatever that is) of two non-orthogonal metrics? In IAYPA, a 1-yard pass on the 50 yard line counts the same as a TD pass to the back corner of the end zone (with all 11 defensive players in the end zone). That ain't right.
(Passing yards + 20 * Passing Touchdowns - 45 * Interceptions) / (Passing Attempts)
= Adjusted YPA (Pro Football Reference)
I don't want to add in touchdowns for several reasons.
First, touchdowns are volatile. Error-prone quarterbacks tend to stay error prone (*cough* Bret Favre *cough*), but touchdown % can vary a lot with the supporting cast. In 2006 Tom Brady threw 24 touchdowns, in 2007 (when Moss showed up) he threw 50. So I'd prefer to set the touchdowns on a separate column.
Second, IAYPA is simple and correlates better with winning than TD/INT or passer rating. So again, put the other indicators to the side as supplemental indicators.
Third, I don't like multifactor models because they're not very transparent. I'd rather have key elements more open for inspection.
For this dataset, though, all three approaches tell *about* the same story, as the conditional formatting shows. Tom Brady is #1 on everything on the two-year data, and Brock Osweiler sucks on ice no matter how you measure.
I took the harmonic just to sort, which I intended as a way to favor players where there was a higher level of agreement among the rating systems. I'm still learning how to use this thing, and would have been better off with the arithmetic in this case. But here's the logic on the harmonic mean. These series all have a median and mean of 3.0:
3 3 3 3 3
1 3 5 3 3
5 5 3 1 1
The harmonic mean for the first series is 3.0, for the second series it is 2.27, and for the third series it is 1.83. The benefit is that it adds in some information about volatility. The less agreement, the lower the harmonic mean. The problem with the way I used it here is that it's correcting the wrong way for a ranking system where 1 is good and 5 is bad. So I didn't get the effect I wanted. On the other hand, the r2 between the harmonic and arithmetic means in this dataset is 99%, so no big deal.
Here's an interesting article that (I think) uses it correctly.
Having acknowledged these errors, I nevertheless stand by my assessment that Tom Brady is a good quarterback, and that Brock Osweiler is bad.
If you wanted to keep it very pure, you could just look at YPA, then sort by INT %. Useful rule of thumb: none of the top 10 in the above analysis have an INT % >2.0. All of the bottom 10 do.
If you're going to factor in the completion yardage at all, then touchdowns need to count for more yards, because they only account the yards to the goal line.
If you really wanted to factor out the "supporting cast," you'd also have to account for drops and yards-after-catch.
Here's the effect this has on me: when I read one of your IAYPA posts, I immediately go to Pro Football Reference to find out what the AYPAs are.
"Inferior people should not be employed." - I Ching
Post a Comment
<< Home