Experience: Something to Build On?
Heading into the 2007 season, the college football writers who vote in the weekly Associated Press Top 25 Poll deemed UCLA the fourteenth-best team in the country, just behind Georgia and just ahead of Tennessee.
Let’s check the resume:
• 2006 Record: 7-6, including home wins over Rice, Utah, Stanford and Arizona;
• Offense: 63rd in rushing, 56th in passing, 71st overall in 2006;
• Defense: 9th in rushing, 87th in passing, 35th overall in 2006.
OK, the offense was sub-par, but the defense was pretty good. There must have been something else generating the Bruin enthusiasm, besides a win over USC the year before… There it is:
• Returning starters: 10 offense, 10 defense, 20 total.
One of the Bruins’ most vocal supporters during the offseason was Sporting News college football writer Matt Hayes, who hedged his obligatory USC-LSU championship game pick with reminders that nine different schools had won the BCS national championship in the past nine years, all the while singing UCLA’s praises.
“The Bruins have 20 starters returning from a team that found itself late last season, holding mighty USC to nine points in the season finale and keeping the Trojans from the national title game,” Hayes wrote. Hayes dismissed the Emerald Bowl blowout loss at the hands of Florida State as a mere “aberration,” which must have been different from the five other 2006 losses.
What did all that experience get head coach Karl Dorrell at the end of 2007? A 6-7 record and a pink slip.
The returning starters theory in college football is nothing new. In a sport where stud recruits likely have a career window of three years, experience is the optimistic fan’s favorite cure-all for the lingering hangover of a disappointing season. For handicappers and pundits, that extra ingredient can transform a bad team into a good one or a good team into a great one.
It actually sounds pretty convincing. Unfortunately for hopeful fans, recent data don’t necessarily bear that out.
And The Numbers Say?
To construct useful measures of the link between returning starters and success, I culled information from all teams in the six BCS conferences, dating back to 2002. (The total number of teams came to 380.) I included returning starters broken down by offense and defense. Additionally, I collected data on the previous season’s winning percentage.
As a baseline, the average team during this period returned 13.23 starters, 6.50 on defense and 6.72 on offense. The average record was 7-5.
I arrived at a few conclusions:
*How your team did last year says a lot more about what you’ll do this year than how many guys you bring back.
For the six BCS conferences, the numbers reveal a correlation close to zero, 0.01, between returning starters and teams’ winning percentage. Not surprisingly, the only variable considered among the general population of teams that shows a strong relationship to performance is the previous year’s winning percentage, which has a positive correlation of 0.60. The numbers also suggest a relatively weak link of 0.20 between the number of returning starters and the change in wins from year to year.
*If you think your team has a shot at competing for a national title, don’t worry too much about having boatloads of playing time in previous seasons.
I drilled down to the top 10 percent of teams by winning percentage in the six-year period in an effort to see if winning teams had more experience. Just like the general population, nothing really stands out among the best teams, as the correlation of returning starters to winning percentage is 0.01.
Notably, the top 10 percent of BCS teams in terms of winning percentage have had an average of 13.05 returning starters: 6.55 on offense and 6.50 on defense. Those figures are equal to or less than all three measures for all BCS teams during the period.
If you’re looking for further proof that experience is meaningless to elite teams, consider the particulars of this group. Four teams had 17 returning starters, the highest number of returnees–2007 USC, 2005 Penn St., 2006 Louisville and 2003 Ohio St. None won a national championship.
The Buckeyes’ case is particularly notable, in that the 2003 squad had six more returnees than the prior year’s national championship team, yet failed to defend their crown. In contrast, the 2004 USC Trojans, arguably the best team of all time, returned just nine starters from the previous season, the lowest of any team among the group.
So the 18 starters Ohio St. and Texas Tech bring back aren’t too significant. On the other hand, USC’s not out of it with just 11.
*Really bad teams can’t blame their lack of experience. But they might be able to blame a wealth of it.
The bottom 10 percent of teams by winning percentage—a group of 41 teams—show negative correlations of 0.19 and 0.16 between returning starters and wins and returning starters and winning percentage, respectively. This has two implications for the worst BCS conference teams. First, the relatively low correlation indicates a weak relationship between returnees and poor performance. More importantly, the fact that these correlations are negative means the data suggests that wins tend to go down among this group as more starters return. This suggests bad teams are hurt to some extent by not turning over their roster the next year.
Duke, this means you’re not out of the woods, even with the 17 guys you’re bringing back. Same goes for you, Stanford.
Turning It Around
What about the teams that improved the most from one season to the next? Somewhat stronger relationships are evident among this group.
The top 10 percent of teams in terms of increased wins gained between four and seven victories between seasons. These teams have a higher correlation, 0.36, than the general population when it comes to the relationship between returning starters and difference in wins. Also, a very strong relationship—0.85—can be found between the previous year and the following year’s winning percentage.
Given the strong correlation between winning percentage in the two years, I tried to build a profile of the most improved teams by working backwards using the data I had collected. I realize this is pretty much statistical alchemy at this point, but whatever.
First, these teams returned a higher average number of starters than the general population: 13.88, with 6.76 on offense and 7.13 on defense. The group had an average of 4.76 wins the previous year, climbing up to 9.49 the following year, with average winning percentage moving up from 0.40 to 0.73.
Then, I created a statistic for this cohort that incorporated both experience and past performance, “history,” which proved to have a correlation of 0.90 to the current season’s winning percentage. History consists of returnees as a percentage of total starters plus three times the previous year’s winning percentage:
Making The Leap
So here’s the profile of the average team among this group:
• History = 2.04
• Previous Year’s Wins = 4.76
• Previous Year’s Winning Percentage = 0.40
• Following Year’s Wins = 9.49
• Following Year’s Winning Percentage = 0.73
• Difference in Wins = 4.73
• Total Returning Starters = 13.88
• Offensive Returning Starters = 6.76
• Defensive Returning Starters = 7.13
Who fits the profile to be a high-riser this year? Here are five teams to think about:
Colorado
• History = 2.02
• Previous Year’s Wins = 6
• Previous Year’s Winning Percentage = 0.46
• Total Returning Starters = 14
• Offensive Returning Starters = 6
• Defensive Returning Starters = 8
Maryland
• History = 2.02
• Previous Year’s Wins = 6
• Previous Year’s Winning Percentage = 0.46
• Total Returning Starters = 14
• Offensive Returning Starters = 9
• Defensive Returning Starters = 5
Louisville
• History = 2.09
• Previous Year’s Wins = 6
• Previous Year’s Winning Percentage = 0.50
• Total Returning Starters = 13
• Offensive Returning Starters = 5
• Defensive Returning Starters = 8
Pittsburgh
• History = 1.93
• Previous Year’s Wins = 5
• Previous Year’s Winning Percentage = 0.42
• Total Returning Starters = 15
• Offensive Returning Starters = 8
• Defensive Returning Starters = 7
Northwestern
• History = 2.14
• Previous Year’s Wins = 6
• Previous Year’s Winning Percentage = 0.50
• Total Returning Starters = 14
• Offensive Returning Starters = 7
• Defensive Returning Starters = 7
Of this group, Northwestern looks pretty intriguing. The Wildcats face an in-conference schedule in which the road games are imminently winnable. Likewise, the non-conference slate looks an almost assured 4-0 start: Syracuse, at Duke, Southern Illinois, Ohio.
The same goes for Maryland. The Terrapins get Wake Forest and Florida State at home during the ACC schedule. Out of conference looks manageable as well: Delaware, at Middle Tennessee, Cal and Eastern Michigan.