Monday, November 23, 2009
Understanding the BCS computers
By Brad Edwards
Special to ESPN.com
For many college football fans, there is still plenty of mystery when it comes to the BCS.
The formula that determines which two teams will play for the national championship was relatively complex in its early days and originally included a statistical element called a "maximum adjusted deviation" that caused many heads to spin.
It's since been simplified to the point that it's now just two parts polls and one part computers. But even that isn't exactly simple to the average observer.
Most fans understand the polls and how they work. The computers, however, can be a major source of confusion.
Therefore, this version of Road to the BCS is an attempt to remove some of the mystery from the computer element of the BCS formula.
After years of trial and error, the BCS administration has settled on six computer ratings systems that have been a consistent part of the formula since 2004. Under the current setup, the role of the computers is essentially to settle the disagreement when the coaches' and Harris polls don't have a clear-cut opinion of which are the two best teams in college football.
So, how do these computers go about rating the teams?
In a very general sense, these six computer programs judge teams primarily on record (wins and losses) and secondarily on schedule strength, giving some consideration to location of games and absolutely no value to scores of games (margin of victory or defeat).
That's an oversimplification, since there are factors that make every computer unique, including each having its own way of calculating schedule strength. In fact, they are so unusual that, over the last six seasons, there has never been a week in which any two of the computers have produced an identical top 10. (The coaches' and Harris polls, by the way, have had an identical top 10 each of the last three weeks.)
The best known of these is the Sagarin Ratings, devised by Jeff Sagarin, which are printed in USA Today on a weekly basis. Like some of the others, Sagarin doesn't share all the details of his formula, but it is predicated on the combination of record and schedule strength and does consider the site of a game when measuring the value of the result.
The ratings that most closely resemble those of Sagarin belong to Dr. Peter Wolfe. His process involves connecting more than 700 four-year colleges through common opponents and assigning a rating to each one based on the probability of its results.
This week, for example, Wolfe has TCU (11-0) ranked ahead of Texas (11-0) because the Horned Frogs have played a schedule that has offered a greater probability of losing than the schedule the Longhorns have faced. That might be contrary to what voters think about a team from the Mountain West compared to a team from the Big 12, but the bottom line is that games at Clemson, at BYU and against Utah are more dangerous, on paper, than the toughest games on the Texas schedule.
In addition to Sagarin's, Wolfe's ratings also will occasionally show similarities to the Anderson & Hester Rankings. Jeff Anderson and Chris Hester devised their system in the early '90s, and they implement a unique approach to calculating schedule strength by incorporating conference strength into the equation.
For this reason, Alabama will more than likely jump Florida and take over the top spot in the A&H rankings next week if both teams win.
Alabama plays 7-4 Auburn this week, while Florida plays 6-5 Florida State. Not much of a difference. But when you consider that the ACC is the No. 6 conference in these rankings, a win over FSU is worth a good bit less than a win over Auburn.
|TCU is ranked ahead of Texas in three of the computer rankings.|
The Massey Ratings, devised by Kenneth Massey, also have a heavy schedule strength influence. In addition, Massey gives slightly more value to games played late in the season than those played early in the season.
One result of this approach is that Oregon (9-2) is ranked one spot ahead of Boise State (11-0). The Ducks get a boost from a schedule that is far more difficult than what the Broncos have played, plus the Oregon loss to Boise is de-weighted a little because it happened in the season's opening weekend.
A fun feature on Massey's site each week is a list of the season's least likely results, based on his ratings. For those who are curious, the No. 1 headscratcher from the FBS this year is Maryland's win over Clemson.
Like Massey, Wes Colley has Oregon one spot ahead of Boise State this week in his Colley Matrix, which is yet another system that places great value on schedule strength. Unlike Massey and most of the others, Colley considers only wins and losses against FBS competition.
That's probably why Texas ranks higher (2nd) in the Colley Matrix than any other computer this week. The Longhorns are the only one of the six unbeaten teams to have done it all against FBS opponents, and the win over Oklahoma State rates better within Colley's system than any win by Florida, Cincinnati or TCU.
That brings us to the final of the six computers, the Billingsley Report. Richard Billingsley is more college football researcher and historian and less mathematician, so his rankings often look a little different than the others.
One example is that he has Boise State ranked higher (6th) than any other computer this week, and his is the only computer with the Broncos ranked more than one spot ahead of Oregon. That's because Billingsley's system uniquely provides a head-to-head advantage whenever two teams that played each other have similar rankings.
He also calculates schedule strength based on the rankings of the teams rather than their records, and, like Massey, he gives more weight to results that happen later in the season.
Right or wrong, this all adds up to Billingsley's rankings more closely mirroring the polls than any of the other five do. And he's aware that his system has plenty of critics, especially in the math community.
"Some people love my method. Some hate it. But no computer is ever going to satisfy everyone," he said. "That's why the BCS uses six of us."
|Despite the victory over Oregon, three computers rank Boise State behind the Ducks.|
Brad Edwards coordinates the college football research for ESPN and is an analyst for "College GameDay" on ESPN Radio each Saturday throughout the season.