Consistent Inconsistency in Football

Posted on August 1, 2006 by


Much of this forum recently has been devoted to the discussion of basketball.  Our book, though, is about so much more.  The first part of The Wages of Wins examines labor disputes, the ability of team to buy wins in baseball, and competitive balance in various sports leagues.  Not until the fifth chapter does the discussion fully focus on the NBA.  This focus, though, does run through much of the remainder of the book.

The exception is Chapter Nine, which is entitled “How are Quarterbacks Like Mutual Funds?” Originally it was titled “Consistent Inconsistency” but Stacey thought this title read better – and he was right.

A casual reading of Chapter Nine might suggest that we are simply seeking to build a productivity model for the NFL much like what we have for the NBA.  And if you read the chapter, a productivity model is what you will find.  A productivity model for the NFL, though, is not the primary story we are telling in the ninth chapter.

Before I get to this primary story, though, let me briefly discuss the aforementioned productivity model.  Via a pair of regressions we presented a method of measuring the number of Net Points a quarterback produces.  We also used this analysis to measure Wins Produced, much like we did for NBA players.  Although the calculation of Net Points and Wins Produced is nice, we also demonstrated that quarterback’s value could be captured by a metric we call QB Score. In an earlier post, QB Score was explained as follows: 

We were able to estimate the relative value of Yard Gained – which includes rushing and passing yards – Plays – which includes passing attempts, sacks, and rushing attempts – and Turnovers – which includes interceptions and fumbles.  Our research indicates that one play – in terms of wins and points – is worth about three times the value of a single yard.   A turnover is worth about 50 yards.  Now these values – 3 and 50 – are not exact.  But it is close enough to give you a quick estimate of a quarterback’s effectiveness.  Given this, QB Score – which is both less complex and more accurate than the NFL’s quarterback rating system — is calculated as follows: 

QB Score = Yards  – 3 X Plays – 50 X Turnovers

Unlike the NFL’s quarterback rating system, QB Score is simple, it is complete, and it appears to be accurate.

For those who are interested, I examined each quarterback who attempted 224 passes in 2005 – the minimum necessary to qualify for the NFL’s quarterback rating leaders.  The results are posted HERE.

The top quarterback in 2005 was Peyton Manning.  As noted in our book, an average quarterback posts a Net Points per Play of 0.133 and QB Score per play of 1.1, and Manning was well above these averages.  The other top quarterbacks in 2005 included the Super Bowl quarterbacks – Ben Roethlisberger and Matt Hasselback – as well as Jake Plummer, Trent Green, Carson Palmer, and Tom Brady.

None of these quarterbacks play for the Detroit Lions, the team I have followed since my childhood in Detroit.  The Lions have moved from Joey Harrington – who was well below average last year – to Jon Kitna and Joshua McCown.  Kitna was slightly above average last time he played on a regular basis in 2003.  McCown was below average last year, but posted better numbers than Harrington.

Now what do all these number mean?  It turns out that these numbers mean very little.

To explain we need to consider a story told in Chapter Seven of our book.  In this chapter we presented evidence that NBA players were fairly consistent from season to season. As we note in the book, in describing something with words like consistent or inconsistent, one must have a reference point.  Relative to what are NBA players consistent?

Answering that question is the purpose behind Chapter Nine.  Around 70% of an NBA player’s Win Score in a given season is explained by the player’s Win Score the previous season. 

As we report in the book, NFL quarterbacks are far less consistent.  With respect to QB Score, less than 15% of what a quarterback does in a given season is explained what he did the previous campaign.

In the book we go to some lengths in examining the issue.  The conclusion, though, is pretty clear.  NBA players appear to be much more consistent than NFL quarterbacks.

With a bit of thought, it is obvious why quarterbacks are so inconsistent. A quarterback’s performance depends upon the actions of ten other teammates, the quality of defense he plays, and the quality of plays called by the coaching staff.  Much of this is beyond the quarterback’s control, so as Brett Favre discovered last year, a poor supporting cast can make even a good quarterback look bad.

That is the basic story we tell in Chapter Nine.  NFL quarterbacks are consistently inconsistent.  So knowing the productivity of quarterbacks in 2005 does not tell us much about what we can expect to see in 2006.  And that means I can have hope as a Lions fan.  Of course, it is August and Lions fans always have hope at this point.  Only in September do our thoughts once again turn to the April draft.

So inconsistency of quarterbacks is the truly important story of Chapter Nine. But why is that important? As detailed in the book, the consistency issue goes back to the link between wages and wins. And perhaps more needs to be said on that in this forum.

But that will have to be for another post. 

Let me conclude that this not the only topic I will be commenting on with respect to football.  One of our links on our homepage is to Football Outsiders – a site that has a number of interesting statistics sure to delight NFL fans.  As we note in the book, Football Outsiders has even more complicated measures of performance than QB Score.  But do these measures tell a different story? The answer to that question will be posted…. well, not today. But it will be soon.

— DJ

Posted in: Football Stories