Conventional wisdom around the blogowebs and amongst professional gamblers is that the greater the number of returning starters, the better the team should be. Take two teams found in the state of Michigan:
Team 1: had an abysmal year in 2008. Replaces quarterback but returns 15 starters. Serious media outlets predict turnaround year and bowl appearance.
Team 2: had a very good year in 2008. Replaces quarterback but returns 15 starters. Serious media outlets predict dropoff year and crappy bowl appearance.
You've probably figured out by now: Team 1 is Michigan, who went 3-9 in 2008, finished 10th in the Big Ten, and missed a bowl game for the first time in 34 years. Team 2 is Michigan State, who went 9-3, finished 3rd in the Big Ten and went to a New Year's Day bowl game.
I have also been chewing on the Wall Street Journal assertion that Offensive Line Experience Can Predict Success. Unfortunately that article is pretty weak, offers no statistics or links thereto to prove its case, and seems to be cherry picking based on anecdotal evidence. What bothers me about it is that if you had a bunch of sub-quality players on the offensive line, and returned all of them, suddenly they'd be magically brilliant and the team would be able to plug in weak or new skill position players and do well. Likewise, if you had a four-year starter (48 or so games), a two-year starter (26 games), and three true freshmen, would your 74 line starts necessarily mean you had an "experienced" line?
It seems to me that returning starters are important but especially at the skill positions and in particular quarterback since, as I found before, passing (NYPPA+) is much more strongly correated with win percentage than running (NYPRA+). So I set out to try to analyze the improvement of teams' win percentage based on number of returning starters.
I took all the FBS teams' winning percentage from 2007 and 2008 from stassen.com and calculated the delta in win percentage. I then took the number of returning starters for 2008 from Phil Steele and ran the regression on the number of returning starters to the delta in win percentage...and found that there is a very weak correlation between returning starters and improvement in win percentage.
As you can see, improvement (as a positive change in win percentage) is least weakly correlated with returning offensive starters. Therefore, betting on a team improving just because of the number of returning starters is a sucker bet.
[I suspect there's a stronger correlation once you factor in the returning quarterback with the returning offensive starters. If I can ever clean out the cobwebs and remember how to properly set up a simultaneous regression using binary inputs, I will check it out].