In this week's Varsity Numbers, Bill Connelly revisits some measures and concepts: Adjusted Scores, Covariance, and momentum (or whatever you choose to call it).
04 Oct 2004
by Ryan Wilson
When Carson Palmer squared off against Ben Roethlisberger yesterday, they had a combined five career NFL starts. Conventional wisdom suggests they will both struggle at times as inexperienced first-year starters, but which one might struggle more than the other may have nothing to do with their physical talents. What's more important is how well their respective teams performed in the past few seasons. That might sound obvious, but it is often forgotten in the discussion of first-year quarterbacks.
Last Sunday, for example, Jerry DiPaola of the Pittsburgh Tribune-Review wrote that "...unless Ben Roethlisberger is Dan Marino or John Elway, the Steelers have a much greater chance of failing than succeeding this season... The reason is that most teams forced to use rookie quarterbacks in the starting lineup for at least six games end up with losing records."
I think DiPaola is onto something, but I don't think he asks the right question (or at least, he's not specific enough with the question he poses). No one disputes that first-year quarterbacks (specifically rookies) routinely struggle as they learn a new offense, game-plan against more complex defenses, and adjust to the overall speed of the NFL. But the quarterback is still just one player on the team, and how theteam performed in the preceding seasons also has an impact on the current seasons wins and losses. He cites only Marino and Elway as posting winning records their first seasons (both were rookies) but what he fails to mention is that in the previous two seasons before their arrival, the Dolphins won 72% of their games and the Broncos won 48% of their games. And while winning 48% of your games may not sound that impressive, consider it is the best winning percentage of the 12 quarterbacks examined other than Marino.
So the real question is: How well do first-year quarterbacks perform given how well the team performed in the previous seasons. In addition to the 12 quarterbacks mentioned in the column, I also looked at most of the current quarterbacks currently starting in the NFL (I should note that I did this primarily to increase the sample size and make the results a little more reliable, and I'm using "rookie" and "first-year starter" interchangeably here). I then looked at how these teams performed in the two years prior to the quarterback's arrival. Of the 21 quarterbacks examined, 12 actually improved on their teams' combined winning percentage from the two previous seasons while nine of the quarterbacks posted worse winning percentages than the previous two seasons combined. The expanded list of quarterbacks that posted winning records as first-year starters now includes Shaun King and Chad Pennington.
Of course it's important to keep in mind why a team starts a rookie quarterback (rebuilding, injuries, contract disputes), and why a team might be destined to lose this season even when they were successful in previous seasons (free agency, poor draft, rebuilding). Either way, here's an extension of the table DiPaola cites in his column:
|STARTS||WIN %||TEAM WIN %
|TEAM WIN %
|(note: FIRST YEAR = first year QB started at least 6 games
TEAM WIN % YEAR -1 = team's winning percentage the year before
TEAM WIN % YEAR -2 = team's winning percentage two years before)
To take it a step further, the correlation between how many games a rookie quarterback wins and how many games that team won in the previous season is 0.38. (Correlation is explained here) And the correlation between how many games a rookie quarterback wins and how many games that team won in the two previous seasons is 0.31. In English, this means that the more a team wins in previous seasons, the more the team is expected to win this season. And while this isn't earth-shattering news, it does explain why Marino and Elway had more success as rookies than, say, Bledsoe and Aikman.
So if a rookie quarterback's success is at least in part reliant on how a well (or poorly) a team performed in the two previous seasons, there's both good news and bad news for fans of Carson Palmer and Ben Roethlisberger.
If you are a Bengals fan, it's certainly good news that the Bengals were 8-8 last season. Unfortunately, they were also 2-14 the year before. Similarly, the Steelers were 10-5-1 in 2002 but fell to 6-10 last season. The two-year winning percentage prior to Palmer taking over was .313; for Roethlisberger it was .516. The Steelers winning percentage of .516 ranks seventh in the table above (the Bengals winning percentage of .313 would tie for 18th) -- and four of the six teams with better two-year winning percentages had their first-year quarterbacks do no worse than .500 (Of the six, only Bulger and Griese finished below .500; Marino, King, Pennington, and Boller all finished with winning records in their first year as starters).
How many games can the Bengals and Steelers expect to win this season with first-year starters? Well, I ran a regression of wins in the current season based on wins in the previous two seasons for teams with first-year quarterbacks, and found that the Steelers' predicted win total is 7.3 games and the Bengals' predicted win total is 6.1 games. Unfortunately, these predicted win totals can be very kindly described as loose approximations (primarily due to small sample size). In the spirit of full disclosure, the 95 percent confidence interval for the Steelers predicted win total this season is between 4.4 and 10.2 games; for the Bengals, their predicted win total was between 4.3 and 7.9 games. (Math geek explanation: This means that, based on the data, I'm 95 percent confident that the Steelers' actual win total will be between 4 and 10 games; for the Bengals, I'm 95 percent confident they will win between 4 and 8 games. Which isn't really that precise, is it?)
But, despite Jerry DiPaola's concerns that rookie quarterbacks are more likely to post losing records, a quarterback's experience probably isn't one of the most important factors in a team's success. Of course, in the NFL, "probably" never means "definitely." It will be interesting to see how both quarterbacks progress this season. If Palmer continues to struggle, does Marvin Lewis turn to Jon Kitna? If Roethlisberger continues to improve, does Tommy Maddox change his name to Wally Pipp? My gut tells me the answer to those questions is "no" and "yes," respectively, but my gut also told me that Heath Shuler would be a good player and that Tom Brady would be awful -- and we all know how that turned out.
(Updated Monday night to include Kyle Boller and fix errors in Denver's record prior to Brian Griese taking over as starting quarterback. Adding Boller also changed correlation coefficients.)
2 comments, Last at 01 Feb 2007, 12:41pm by Amatory