A new study of 25 years of data from a major college football poll challenges three strongly held beliefs of many coaches and fans.
Contrary to conventional wisdom, the research found that teams are not punished by pollsters for losing late in the season; that there's no benefit to playing and beating strong opponents; and that the margin of victory doesn't matter in poll results.
“In the end, what the study really says is that voters in college football polls take into account winning and losing – and that is about all,” said Trevon Logan, author of the study and assistant professor of economics at Ohio State University.
“A lot of the things that coaches and fans believe turn out not to be true.”
Logan's paper is available at the website of the National Bureau of Economic Research (www.nber.org/papers/w13596 ).
The study is unique in the size and scope of its data set, Logan said. With the help of about 10 undergraduate students, Logan assembled the weekly Associated Press poll results for 25 of college football's most prominent programs for the 25 seasons from 1980 to 2004. In all, the data set contains information on more than 6,000 games. Logan examined the number of points teams received from the AP voters each week. Teams are then ranked by the number of points they received.
Logan's data set contains the date of the game, location, final score, opponent, opponent's record at the time of the game and at the end of the season, and the team's and opponent's ranking before and after the game in the AP poll.
One of the more controversial findings has been that teams are not punished for losing late in the season, Logan said. In one sense, losing early does help, he said. Teams that are beat at the beginning of the season have more time to make up lost ground as teams higher in the rankings lose.
But Logan said his findings show that pollsters do not punish a team that loses a late-season game by moving it further down in the rankings compared to a team that loses early. In fact, the exact opposite is true.
Logan defined late-season losses as those that occurred in the tenth poll week of the season or later, which usually comes in the last week of October or the first week of November. He found that the cushion provided by losing late in the season is around 20 percent of the value of losing. The results suggest that more than three-fourths of AP poll voters rank a team one place higher in their rankings after a late season loss than they do for an early season loss.
The study also found that beating a strong opponent – one that has more wins and a higher ranking – did not help a team's ranking any more than beating a weaker opponent. However, Logan did find that losing to a strong opponent actually softens the blow of a loss. For example, losing to a team with an 8-3 record would actually decrease the negative point change from losing by more than 15 percent of the change for an average loss.
The study also rejected the belief that teams are helped by blowout wins. Results showed that big margins of victory – defined as more than 17 points – did not give teams an advantage among pollsters when compared to much closer wins.
While blowout wins don't help, it does hurt teams when they suffered a blowout loss. A team loses about 20 percent more points in the AP poll if they lose by a wide margin as compared to a small margin. On the other hand, close losses help – they reduce the reduction of points from losing by about 10 percent.
So why is the conventional wisdom about college football polls so wrong? Logan gives several reasons.
One of the most important is that many coaches and fans insist on using one or more anecdotes to establish their claims, he said. Many fans can point to one or two games in particular where the conventional wisdom turned out to be true – all the while ignoring the many other cases that contradict it.
“There's a standard practice of taking one or two instances and making it into a general rule,” Logan said. “But anecdotes aren't evidence.”
This is especially true because fans tend to fixate on the very top three to five teams and believe whatever happens to them is a general rule that happens to all teams.
“Perhaps due to its rich and colorful history, college football gives rise to stories and anecdotes that become ‘truths' without any careful consideration,” Logan said.
While the current system of naming a Division I college football championship has created a lot of controversy, Logan said this study suggests that the playoff system proposed by some critics would not solve the problem.
If any system were to determine the final four or eight teams to compete in a playoff, it likely would include human polls like the AP poll in this study. (During part of the time period covered by this study, the AP poll was used in the Bowl Championship Series formula which determines the teams that play in the national championship game. While the AP poll is no longer a part of the formula, a similar human poll has taken its place.)
“The results here suggest that losing late in the season, for example, could help some teams and have implications on who would make the playoffs, as long as rankings were based on human polls,” Logan said. “The stakes of these cutoffs are potentially quite high.
“If you're going to only take the top 4 or top 8 teams in a playoff, you had better make sure they are the best teams,” Logan said. “We have to think about the biases in the polls that will affect who gets in.”
Source: Ohio State University
Explore further: Australia probes spy case at top science authority