Is the BCS killing the Bowls?
Is BCS killing the Bowls?
Good background article on BCS computer polls.
NY Times - Who Programs the Computer Polls
In the College Bowl Race, the Crucial Players Are the Programmers
By COREY KILGANNON
Published: December 4, 2003
HIS weekend will mark the end of the regular season in college football, and barring upsets of the top-ranked teams, there will be a tight race for the No. 2 spot in the nation.
So who would then help determine which team - Louisiana State or Southern California - would play top-ranked Oklahoma for the national championship, and which would be consigned to a lesser bowl game?
Advertisement
Why, an astrophysicist, of course; and an immunologist and an M.I.T.-trained mathematician from Indiana, not to mention a math professor from Virginia.
Granted, this is only part of the puzzle, but a crucial part nonetheless. Although games are won and lost on the field, the big-picture results come well after the last interception, fumble or field goal, when rankings derived from elaborate computer formulas are factored into the race known as the Bowl Championship Series.
Each week, seven PC's - scattered in various bedrooms, living rooms and offices around the country - calculate rankings of the nation's teams after being fed the same game results. Each week, each computer ranks the teams differently. And that's when the arguing begins.
Ask the fans of Georgia - No. 5 in the polls of sportswriters and coaches, but brought down to No. 7 by its lower computer rankings. Or figure the computer science that elevates Miami of Ohio to No. 4 in two computer rankings but leaves it at a more terrestrial No. 23 in another.
What comes out of the computers, it seems clear, depends on what's being put in. But you won't get much help from the computers' operators in divining what that is.
"Now you want to look under the hood of my car," said Jeff Sagarin, whose computer rankings have been published in USA Today since 1985, deflecting an inquiry about his own formula.
Computers have been cranking out football rankings for a generation - The New York Times has published its version for college teams since 1979 - with the notion of injecting some objectivity into a weekly ritual otherwise based on polls of sportswriters and coaches. For most of that time (and for pro teams, even now), it has been merely an intellectual exercise. But since the advent of the Bowl Championship Series in 1998, the computer rankings have been a major component, along with the polls, in determining the matchups of the top four bowl games - and the split of about $90 million.
Fans (and even officials of the conferences involved in the B.C.S.) know very little about exactly how these computers make their choices, where they are based and who is behind the programming. As it happens, the seven computers used in the bowl race are not run by large institutions or organizations (except for this newspaper), but by mathematicians and statistics geeks who also happen to be sports devotees.
The computers make their selections not just on victories and losses, but also on such peripheral matters as the records of a team's opponents. They are based on the same statistics, but use them differently to assess teams, assigning different priorities to different pieces of data - considering whether a team was playing at home or away, or weighing its recent games more heavily than earlier performances.
Take Dr. Peter Wolfe, who has a system employed by the B.C.S. that uses what he calls a "maximum likelihood estimate" obtained partly from comparing how teams performed against common opponents. His system gives no importance to whether a team plays at home or away and weighs all games equally, regardless of when in the season they are played. That's about all he will reveal about his recipe.
"The more you specify," he said, "the more you annoy the readers."
Mr. Wolfe, 49, is a Harvard-educated immunologist in Beverly Hills, Calif., who specializes in AIDS research. But his weekends belong to college football.
He grew up in Pittsburgh during the heyday of the Steelers and attended Pitt in its football glory days. His love of sports statistics drew him to Strat-O-Matic Baseball, a board game based on the records of real-life players. He began doing computer rankings in 1984, writing a program on an early Macintosh using the Basic code he learned in high school and a few self-help books.
As his rankings have become part of the Bowl Championship Series calculations, he has become used to vast amounts of nasty e-mail about his results. "I don't take it personally," he said. "It's understandable."
And as if to prove that programming is an art, some rankers produce multiple versions of their work, with varying results.
Mr. Sagarin, for example, has a set of rankings called Predictor, which he considers his most authoritative. But they emphasize a team's margin of victory - a factor disallowed since last year in the bowl race, to avoid giving teams an incentive to run up the score. So he supplies an alternate version.
Advertisement
"The B.C.S. just wants wins and losses, rankings that look the best when you're looking back at the teams' final records," he said. "But the old-timer beer-and-cigar sports fans want to use the rankings as predictors, so they look at point spread because there's a difference between winning 3-0 and winning 79-0."
Mr. Sagarin, a 55-year-old statistician who lives in Indiana and who majored in math at the Massachusetts Institute of Technology, has worked as a statistician for the Dallas Mavericks pro basketball team, devising a computer program to calculate which lineups produce the best results and determining the best free agents to sign.
He began using computers to rank teams in 1972, adapting his program from a system used for chess players. To arrive at his formula, he said, "I basically look at who you played, where you played them, and the game result."
Not all rankers emphasize the home-away component, but Mr. Sagarin does. "It's a huge advantage to play at home," he said. "Playing an away game is like running a sprint with a grand piano on your back."
So how does he allow for it in his rankings? "Let's just say I factor it," he said with a chuckle.
Kenneth Massey, 27, a math instructor at Hollins University in Roanoke, Va., whose rankings are also used in the bowl series, says his system basically assesses each team on its record and its opponents' strength. And like other computer systems, he said, "it has no ego, so it has no problem correcting itself week to week, unlike poll voters, who might form loyalties.''
The only computer rankings used in the bowl race whose formula is made public - in fact, posted online at
www.colleyrankings.com - are those of Wesley N. Colley, 32, a University of Virginia professor who works on missile tracking technology for NASA. But unless you are a mathematician - or have a doctorate in astrophysical sciences from Princeton, as Dr. Colley does - you might have some trouble parsing it.
"It's basically a linear equation with 117 dimensions," he said. "The program analyzes the data until it comes up with the self-consistent solution. Solving it by hand for one team would take a year."
Dr. Colley said that his approach, which modesty does not prevent him from calling the Colley Bias-Free College Football Rankings, "adjusts for strength of schedule, but without bias toward conference, tradition or region."
With a résumé that lists not only academic work at Harvard but also the 19 stadiums where he has attended major college football games, Dr. Colley acknowledged, "It's a funny marriage, the football fan and the astrophysicist." He defended the computer rankings - "at worst, the computers provide a sanity check against the pollsters; at best, they're telling you something objective and counterbalancing the humans" - but he is certainly aware of their critics.
"People e-mail me saying, 'How much are they paying you?' " he said. "The answer is zero. I can't even get the free tickets to the games." (For the record, he's a Georgia fan, though his own rankings put the Bulldogs at No. 9, lower than most others.) Not every team's fans, of course, would resort to sending e-mail to take issue with a computer. Some pick up the telephone.
"Nobody ever calls to say they're happy with the week's rankings," said Marjorie Connelly, who oversees the rankings at The New York Times. "People calling seem to think there's a human interaction in the system, that someone at The Times makes these picks. It's not like, once the season starts, we see a team in our rankings and say, 'Ooh, we have to fix that.' "
The Times system, created by sports editors and computer experts, initially used a mainframe computer, though it now it runs on a Compaq PC sitting on Ms. Connelly's desk. She said the system had changed little since its inception, with results "factored slightly" for home or road appearances and more importance given to games played later in the season. ("If you're going to lose," she advised, "lose early.")
Ms. Connelly, a staff editor in The Times's news surveys department, also conducts political and public opinion polls for the paper and is not a football fan. "I have no vested interest in who's No. 1," she said, "just that it's done properly."
Ranking teams by computer "is as much an art as it is a science," said Jerry P. Palm, 39, of Schererville, Ind., an independent computer analyst who runs a Web site called collegebcs.com, which keeps tabs on the computer rankings and publishes standings.
Mr. Palm said that in addition to Dr. Colley's, the other computer formulas should be open to scrutiny. "As it is now, nobody knows the other six criteria," he said, "so how can you know if they make a mistake?"
Meanwhile, heading into the season's final weekend, Dr. Colley has a question of his own. If his laptop crashes, he said, "what'll happen to L.S.U.?"