November 26, 2007

David Fox
Rivals.com College Football Staff Writer

? MORE: BCS standings | Coaches poll voters | Harris poll voters | Huguenin's weekly analysis

Eight sets of rankings are used in the BCS standings: The USA Today coaches' poll and Harris Interactive poll each make up one-third, while the final third is an average of six computer polls.

In recent years, the BCS rankings have undergone several tweaks. The BCS replaced the Associated Press poll with the Harris poll, put greater weight on the human polls, removed strength of schedule and decided to use only computer polls that don't consider margin of victory.

Here's a run-down of each component.

THE COACHES' POLL

Sixty Division I-A coaches rank teams one through 25 weekly in the coaches' poll, compiled by USA Today.

The 60 coaches are selected by the American Football Coaches Association and reviewed by USA Today (Click here for the list). The only coaches not allowed to vote are coaches of teams on major probation (those teams also aren't eligible to be ranked). The AFCA urges all coaches to make themselves available for the poll, but some decline to take part (usually first-year coaches). There are no tenure requirements.

About a quarter of the voters change each season, either because coaches decline to take part or no longer are Division I-A coaches.

For fear of week-to-week scrutiny of the ballots, coaches' ballots are kept secret. Only the final-week ballots are released to the public.

The coaches vote in a preseason poll, and then vote weekly throughout the season.

THE HARRIS POLL

In replacing The Associated Press Top 25, which removed itself from the BCS formula after the 2004 season, the BCS turned to the Harris Interactive Poll.

Harris Interactive randomly selects voters from among the more than 300 people nominated by conference offices and Division I-A universities.

Unlike the coaches, Harris poll voters do not submit votes until Sept. 23. The coaches participate in a preseason poll and rank teams from the first week of the season.

Before the season, Harris Interactive releases the names of the 114 voters, but does not make public any information that describes the voters' location or connection to college football.

Like the coaches' poll, each voter's final ballot will be made public.

THE COMPUTERS

Six computer rankings make up the final third of the BCS formula. The best and worst computer rankings are thrown out, and the sum total of the remaining four is divided by 100 (the maximum possible points) to come up with the BCS' computer rankings percentage.

The computers are included with the intention of removing human bias.

Though the BCS removed strength of schedule as a stand-alone component of the standings in 2004, schedule strength is included in some form in each of the computer rankings. The BCS also removed margin of victory from the formula, and uses only computer polls that do not take final score into account.

Each computer ranking has its own formula, using different ways to calculate strength of schedule and, in some cases, conference strength.

Here is a brief description of each computer ranking:

The strength-of-schedule component judges the records of each team's opponents and opponents' opponents.

The rankings also put weight on conference play in strength of schedule. Conference strength is rated by each league's non-conference record and difficulty of the league's non-conference schedule.

The Anderson & Hester rankings also take into account home and road records in evaluating strength of schedule.

This computer poll takes team's opponents' rank into account in strength of schedule rather than simply a team's record. For example, a team would get more value from defeating Arizona than Buffalo, though both teams might be 4-6.

The Billingsley Report also claims to be more sensitive to head-to-head competition for "comparable" teams. For example, that means Wisconsin would be ranked ahead of Michigan since the teams had identical records and similar strength of schedule following the Badgers' win. But despite its win over the Buckeyes, 9-3 Illinois would not be ranked ahead of 11-1 Ohio State.

"It's a not a complicated system; it's just simple math with some powerful rules," says Richard Billingsley, founder of the report.

Strength of schedule has a strong influence on the final ranking, but it is calculated by a simple won-loss record. In the formula, a four-loss team playing against opponents with better winning percentages will be ranked ahead of a two-loss team facing weaker opponents.

The rankings claim to be free of bias toward conference, tradition or region. All teams are assumed to be equal at the beginning of the season, so Michigan starts at the same point as Eastern Michigan. As a result, early polls bear little resemblance to the human polls.

This computer model does not account for home and road wins.

Colley won't be surprised if his rankings differ from the AP and coaches polls, the two most commonly cited rankings.

"The coaches and AP pollsters agree with each other about four times better than they agree with me," Colley writes on his Web site. "However, one would expect that the coaches and AP polls agree very well for a very simple artificial reason. The coaches read the AP poll and the media voters read the coaches' poll, so there is statistical feedback between the two polls, right or wrong. A computer scheme, like this one, does not read other polls."

BCS computers are not permitted to use margin of victory, but points scored and points allowed are a factor for Massey, including calculations for a team's home-field advantage. The site of the game is included in the formula for schedule strength. Preseason ratings based on each teams' postseason ranking is included, but that effect is "damped out completely" by the end of the season.

Though margin of victory is not a factor, the final score is used to calculate the probability a team would win a rematch under the same conditions. Massey indicates that the scores help assign an objective value to non-quantifiable elements, such as motivation or weather. A team that wins a close, high-scoring game receives a lower probability to win a rematch than a team that wins 10-0.

In his formula, there are diminishing returns for a team that runs up the score. The difference between a team winning 30-0 and 56-3 is minimal, though there is an advantage to winning comfortably.

Conference strength does not play a role, but non-conference records factor into the rankings; therefore, conferences with better non-conference records will have stronger schedule strength. "A rising tide lifts all boats," Massey says in his description.

The BCS uses his rankings that do not include a margin-of-victory component. Input in Sagarin's rankings for the BCS are wins, losses and site. The site is sorted by home, away, neutral or a "close-by," which would be an LSU game in the Sugar Bowl (but not Florida-Georgia in Jacksonville or Oklahoma-Texas in Dallas, which are neutral-site games).

Since the ratings compound the record of team's opponents and opponents' opponents, strength of schedule is implicit, though it is not considered a separate entity in the formula.

Before the BCS releases its rankings, Sagarin removes preseason rankings from the formula so each team starts on equal footing.

Road wins receive more weight, as do undefeated teams, meaning Division I-AA Northern Iowa ? an 11-0 team with six wins on the road, including one at I-A Iowa State in September ? was 12th in the rankings he submitted to the BCS last week. In the Sagarin rankings that include margin of victory, Northern Iowa was 41st.

Like the rest of the BCS computers, Wolfe's do not take margin of victory into account. "Running up the score is generally looked on as evidence of bad sportsmanship, behavior which should not be encouraged or rewarded," Peter Wolfe, a researcher at UCLA's School of Medicine and the rankings' founder, says on his Web site.

Since the formation of the BCS in 1998, the Wolfe rankings picked the correct national champion six of nine times in its pre-bowl rankings, the exceptions being 2003 LSU (Oklahoma), 2004 USC (Oklahoma) and 2006 Florida (Ohio State).

David Fox is a national college football writer for Rivals.com. He can be reached at dfox@rivals.com.