Sunday, August 19, 2007

Quant We Dump This Statistic?

Mike Reiss' NFL Notes are the best of all the Sunday notes columns presented to Boston newspapers-by a wide margin. Mike is squeezing useful blood out of a turnip of a genre that for the most part the Internet has made a redundant waste of precious newshole space.

However, the lead graphic for Mike's column today was a helpful reminder, in case your portfolio wasn't bringint it to your attention, of the pernicious pitfalls of quantitative analysis-or why sports needs fewer commonly used statistics, not more. One we could certainly do without is the career winning percentage for starting NFL quarterbacks.

This statistic did not exist until a few years ago. No one inside or outside pro football thought it was necessary. Most of us were still grappling with the math of the league passer ratings, wondering how in the world anyone could construct a numerical system were perfection is expressed as 158.3.

Quarterback won-lost record entered the football dislogue because for reasons that escape comprehension, someone in authority felt a need for a new way to express the revolutionary concept "That Tom Brady, he is sure is good!".

As Don Criqui spent much of last Friday night saying, Brady has the best career winning percentage of NFL quarterbacks with over, well, I forget how many starts. It's true. Not instructive, mind you, but true. All that does is get us back to the eternal chicken-egg conumdrum of which came first, the quarterback or his team, which always boils down to opinion, not facts.

Reiss' chart actually proves its own absurdity. All one has to do is remove Brady from it, and look at the win-loss records of all the other NFL quarterbacks. Take a gander, for example at the quarterback right behind Brady in the rankings, Rex Grossman of the Chicago Bears.

Grossman's career record is 17-5, which is splendid. Half the people in Chicago will tell you Grossman's 2007 record needs to be 0-0. He could very well lose his job to Brian Griese for God's sakes. Grossman's record is a reflection of playing for an excellent team. It has only a tangential relationship to his individual contribution to said team.

Brad Johnson has a higher career percentage (and a Super Bowl win) than Carson Palmer. Wanna trade them? Dallas sure would. Palmer's 25-20 record is an indication not of his merit, but of what usually happens to quarterbacks who are number one draft picks, a gruesome rookie season of trial by fire with a lousy team. Take away Peyton Manning's 3-13 rookie year, and his 92-52 record gets much gaudier. On the other side of the QB bell curve of life, Brett Favre's winning percentage, while still good, has been on a steady decline as have his skills and the Packers in general.

The winning percentage statistic for quarterbacks is an example of the "we more graphics in this broadcast" statistic. It's useless. Any baseball numbers nut knows winning percentage isn't close to the best metric for evaluating starting pitchers. Why should football, a more complex and fluid game with infinitely more variables, rush to embrace an incorrect way of examining itself.

Anyone wishing to call Brady pro football's best quarterback doesn't NEED new stats to make the case. He excels by all the conventional measurements of quarterback greatness. He completes a high percentages of his passes for a great many yards, throws far more touchdown passes than interceptions, and has won championships. He's excelled in the clutch. Brady is the most important player on one of history's top teams. He's an all-time great. We don't need another statistic to clutter up that fact.

P.S. Here's the really weird thing about quarterback winning percentages. The all-time leader, and he'll never be touched is Otto Graham of the Cleveland Browns. In the six years Graham was in the NFL, his record was 58-13-1, a percentage of .812. Throw in the four years the Browns were in the All-American Football Conference, and Graham's record soars to an even .850. Yet when asked to name the top 10 NFL quarterbacks ever, Graham's is the name most people forget.

When Graham played, and in the years immediately his retirement, his beyond comprehension winning percentage was used not to praise his individual accomplishments, but to denigrate them. To wit: So what? Anyone could win with that team!

If a statistic can be used to rip the person who leads it, that's a pretty good indication said stat doesn't mean too much.


At 12:16 PM, Anonymous Anonymous said...

This weeks issue of Sports Illustrated (July 2009) states that Daryle Lamonica has the second highest winning percentage of .794 behind Otto Graham


Post a Comment

<< Home