March 25, 2017

More on the quarterback data

So WDIJ put in touchdowns like that inferior website.  There are two reasons.  The first is that I have less certainty that TDs reflect quarterback skill in the same way that interceptions do, the second is that adding TDs increases complexity, but doesn't really change the rankings much.  Here are the two measures plotted against one another:

We know that YPA correlates strongly with winning in the NFL, so that's important.  We know interceptions are costly, and some quarterbacks throw a lot while others throw just a few, so we should adjust for that.

But I'm not convinced that throwing a lot of touchdowns is in the same category in terms of information add.  "Sure, you dummy," I hear you say, "that's why the coefficient's 0.2."  Well ok, but let's put Marshawn Lynch on the Packers for a moment, and let Rodgers hand the ball to him in the Red Zone.  Rodgers' touchdowns and Adj YPA will go down, but I don't think that makes him a worse quarterback.

So if I stick with IAYPA, who am I slighting?  I took each player's rank on IAYPA and compared it to their Adj YPA rank.  Here are the big changes:

IAYPA likes better than Adj YPA

  • Alex Smith (four ranks higher - #10 on IAYPA vs. #14 on Adj YPA)
  • Kapernick, Tannehill, Cutler, and Bradford are all three notches better on IAYPA, but stay in the same general zone.
Adj YPA likes better than IAYPA
  • Cam Newton (four slots higher)
  • Andrew Luck (three slots higher)
That's about it.  Everyone else rates within one or two slots on either metric.

Since all of these guys are in the same general zone, the argument becomes:  which is more important in a mediocre quarterback, the ability to drill the ball into the end zone, or the ability to move the chains without throwing it to the other team too often. 

But I also agree that if I am using IAYPA, I ought to take account of touchdowns in some way, because there's clearly some skill involved, and touchdowns have some passing relevance to the outcome of the game.  ("I always thought if you're kicking field goals you're on your way to losing the game." - Steve Young).

So what can we do about this?

I think I would, if I were not about to ignore the NFL for the rest of my natural life, do something like this.

Historians are somewhat unclear on who first said "three things can happen when you throw the ball, and two of them are bad," but it was probably Darrell Royal.  The problem with this phrase, which has been endlessly invoke by bad coaches with boring teams, is that it recognizes one kind of optionality (interceptions) while ignoring another kind (touchdowns).  In point of fact, four things can happen:

  • The ball is not caught
  • The ball is caught by the other team
  • The ball is caught for a yardage gain
  • The ball is caught for a touchdown and everyone runs off the field cheering and throwing high-fives
It seems irresponsible to ignore that last one.

I know the TD/INT ratio is a sacred to some, but for me it's not useful.  Too much information gets destroyed.  If Brady throws 40 touchdowns with 10 INTs, that is not the same as if Alex Smith throws 12 and 3 in the same number of attempts.  So let's look at the data - who has the most positive optionality and the least negative optionality?  Oh Lawd...

So, there you have a pretty good supplemental indicator, I'd say.  Just a few observations:
  • Remember, this is everyone who played steadily over the past two years, plus Prescott.  No one stays on the field with an INT rate above 3% (Fitzpatrick lost his job), or a TD rate below 3%.
  • Northwest is good, southeast is bad.
  • Over past two years, Brady has been a rage-beast from the 24th dimension.  When the ball leaves his hand, as it often does (#14 in attempts despite missing games), it is six times more likely to end up in the end zone than in the hands of the other team.  So, ok, that does appear to be valuable.
  • Omar...I mean comin'.  Rodgers is the one guy I would say IAYPA clearly has wrong, but Adj YPA doesn't fully correct the error.  This chart shows why he's valuable even with an average IAYPA.  And remember, he's getting a lot of shots of on goal, since he is #6 in attempts over this period.
  • Flacco is almost a coin flip, but his IAYPA is also bad, so we didn't think he was valuable anyway. 
  • Osweiler's a coin flip, but his IAYPA is also bad, so we didn't think he was valuable anyway. 
As a result of this, I will pay more attention to TDs, but not add them to IAYPA.  For me avoiding interceptions is the more important skill, and the one more reflective of quarterback value.  But quarterbacks with average IAYPAs who are northwest on the chart above (assuming comparable attempts) are more valuable than those who are not.

So I propose this summary chart:

Interestingly, the man who emerges as the closet thing to Brady (NE quadrant) is not Rodgers, but Wilson.

Now, you might not agree with my techniques, but if my system is flawed, that would imply that the results are flawed, and that would imply Russell Wilson is not one of the best quarterbacks in the NFL.  And I am not going to sit here and listen to you badmouth Russell Wilson.


Blogger VMM said...

I salute your spirited defense of your eccentric football metrics.

March 25, 2017 at 8:14 PM  

Post a Comment

<< Home