I just finished reading an excellent book titled the The Wages of Wins by David Berri, Martin Schmidt, and Stacey Brook. The authors are three professors of economics, so the statistical analysis is pretty dense and not for everyone. The book discusses why fans don't care about labor disputes, why the NBA teams with the highest salaries don't have the most wins and NBA star players don't help sell tickets to home games, and why competitive balance has never been better in MLB and Bud Selig's Blue Ribbon Panel which found otherwise is wrong.
For the NFL they focused on the quarterback and specifically Brett Favre. They assign a point value to every action that takes place during a game, for example the net impact of an interception is 3.7 points. This is an attempt to create a better measure of a quarterback than the awful NFL quarterback rating. What they discovered when looking at quarterbacks from 1995 to 2004 was that the only thing consistent about them is their inconsistency. They use Favre as an example of this:
"When we look at net points per play, we see that Favre was a top-ten quarterback in five seasons - 1995,1996,1997,2001,2004 - and not ranked in the top ten in six seasons - 1998,1999,2000,2002,2003, and 2005. In fact, according to net points per play, Favre followed his best performance-offered in 2004-by his worst performance in 2005. Inconsistency is the story..."
Next they looked for consistent evidence of inconsistency and discovered that only 10% of a quarterback's current performance can be explained by his past performance. This isn't the case in all professional sports because in the NBA, 66% of current production is explained by his past.
"What do we learn from this simple exercise? It is beginning to look like Brett Favre is not an anomaly. If all we knew was a quarterback's past performance we would not seem to know very much at all about what he would do during the current campaign. This is especially true with respect to turnovers. We found turnovers to be very important in our efforts to explain how many points a team scores and allows. Yet if we looked at a quarterback's propensity to give the ball away in the past, we don't seem to be able to learn anything about his likelihood to commit such offenses in the future. With respect to both interceptions and fumbles, about 1% of current performance is explained by last year's production. For fumbles lost, the explanatory power is only 0.1%. So turnovers, a key facet of a quarterback's performance, seem to be entirely unpredictable.
The lesson to be learned about Favre is that no matter what he has done, or not done, over the past couple of seasons, it is completely unpredictable how a quarterback will play from season to season, or even from game to game. All that matters is what he can do in 2007 and how good his teammates can be. These authors have a great website called the
Wages of Wins Journal to compliment their fantastic book.