How QBs are like mutual funds explained
Posted by disappointmentzone on 1 July 2006
Yesterday I wrote in a post that NFL quaterbacks are inconsistent and I provided a cursory explaination of how the traditional measures of a quaterback’s success–passer rating and win-loss record–are incomplete and, in the case of passer rating, very hard to calculate. A few people sent me an email asking for a more detailed explaination, which I’ll try to provide here. But I want to note that the root of everything I’m about to write is firmly planted in the Wages of Wins, and book worth reading if you are interested in sports and statistics.
There is one primary problem with the passer rating statistic the NFL uses to measure a quarterback’s performance (there are a number of secondary problems as well), which is that passer rating is an incomplete measure of how a quarterback contributes to his team’s success. Passer rating doesn’t account for rushing yards or fumbles. Fumbles are just as costly as interceptions–which are figured into passer rating–and given how passer rating values completion percentage, quarterbacks are rewarded for taking sacks instead of throwing the ball away or attempting to complete a pass by throwing the ball downfield. So passer rating offers an incomplete representation of what a quarterback does to contribute to his team’s success.
A better measure of quarterback performance if a new statistic called QB Score, which takes into account all yards that can be attributed to the quarterback–yards from passing and rushing and sacks–and gives equal weight to both fumbles and interceptions. QB Score can be calculated by following this formula: QB Score = Yards – 3(plays) – 50(turnovers). For reference, a QB Score of 40 is average for a game. So QB score is an improvement over passer rating as a statistical measure of a quarterback’s success.
Of course, both passer rating and QB Score are poor measures for predicting future success. One can only explain 11% and 9.5% of current performance from past performance if one looks at passer rating or QB Score, respectively. Which means that if you know what a quarterback did in one year with respect to either of these statistical measures, you won’t be able to make that good of a guess as to what his future performance will be. The explanatory power of passer rating and QB Score are not high. But no common statistical measure is much better than either passer rating or QB Score, and many are worse. For instance, interceptions per pass attempt has an explanatory power of only 1.1%, fumbles lost per play is only .1%, and passing touchdowns per pass attempt is only 6.5%. Passing yards per passing attempt has an explanatory power 16.5% and rushing yards per rushing attempt, at 26.7%, has the highest percentage of explanatory power. What does this tell us? That quarterbacks are often wildly inconsistent from season to season (and often from game to game).
Why? Because much of what a quarterback does is dependent on how other players perform. Even the greatest quarterback would look like Tim Couch if his offensive line couldn’t block, his running back couldn’t gain yards, and his receivers couldn’t catch. Given how interdependent football is, assigning win-loss records to quarterbacks is a peculiar practice. The average quarterback is responsible for less than 20% of his team’s wins. In 2004 the Philadelphia Eagles won 13 games and Donovan McNabb was responsible for 3.9 wins, about 30% of the total. In the same season the New England Patriots won 14 games but Tom Brady only produced 2.6 wins, less than 20% of the team’s total. Quarterbacks are more responsible than any other player of team victories, but cannot be given full credit for victories or full blame for losses.