The BCS – intelligence that's still too artificial
I’ve been thinking about the practicality of the BCS (Bowl Championship Series) system used to determine the ranking for college football teams in the NCAA. More specifically, I’ve been wondering how true computer science folks feel about it.
The goal of the BCS was to solve a problem that plagued the playoff-less NCAA Division I for years – that rarely would the #1 and #2 teams face off in a championship game to definitively award the nation’s best varsity club. So an algorithm was developed to rank and eventually schedule teams to play each other. But as we’ve seen, it’s been the subject of more criticism than praise.
As a sportscaster, it’s become an easy target from my colleagues in that industry, being the subject of ridicule and recipient of countless finger-pointing for why college football is so screwed up, and why this season let so many people down in accurately crowning a national champion. But as a software developer, my code-writing brethren praise it for its no-mistake, calculated-without-error determination of the best team in the country. The final returned ranking of the nation’s top 25 teams is completely devoid of emotion, politics, sympathy, school loyalty or ambiguity.
But this systems-oriented approach doesn’t take into account the main theme of athletics – that people can go beyond themselves, shatter all expectations and achieve the impossible. And this is primarily why the static model of the BCS is so dangerous to competitive sports – it doesn’t believe in miracles.
It’s natural that most sportswriters would hate the BCS, and conversely that most programmers would love it. It is at its core a decision support system intended to calculate who should be #1 based on a series of criteria like overall record and relative strength of schedule. So, I hear arguments for and against the BCS from both camps.
But as we’ve learned this season, you need a bit of both schools of thought. Just ask USC, who eventually won a share of the first split national championship since 1997, but got screwed out of appearing in the Sugar Bowl by the BCS, as it ranked the Trojans #3 in the country. This was due largely to the fact that the BCS determined that USC’s schedule wasn’t as strong its contemporaries.
Still, using the Coaches Poll – an opinion survey based on the aggregate expertise of those actually involved in the games – USC was voted as the nation’s top team after beating Michigan in the Rose Bowl, and given at least a share of the title as college football’s top dog. But it could be worse. At the moment, the various polls give a weighted consensus on who’s the best in college football, with the BCS carrying the most weight.
Then in the Sugar Bowl, LSU beat Oklahoma (effectively having #2 upset #1, according to their BCS ranking), and won the other half of the national championship after feedback from the Associated Press Poll, another survey based on human input. Think this is a freak occurrence? Ask Miami, who got royally ousted by the BCS in 1999.
We can stand to learn a lot from this example for business – mainly that we can’t rely solely on computer-generated intelligence yet…it’s too fallible. With apologies to Asimov and the Wachowski Brothers, machines aren’t ready to take over the world, and it’s our own fault as those who programmed them. You can’t blame the array of computers that generate the BCS ranking roster. They’re just running the algorithms they were given and returning a resultset, like they’re supposed to. In that regard, the BCS is running perfectly.
It’s a dictum of business that systems are created to not only automate a lot what would be manual, repetitive work, and to make decisions for real people that might add a layer of subjectivity that would corrode an otherwise carefully-calculated product. But it’s also a popular theory that advanced systems exist to eliminate something that has the potential to cause catastrophe in productivity: human error.
As we’ve seen, human intervention has proven to be a good thing. This tragically is something that in so many MIS classes is quite often taught, but in the real world so rarely practiced.
Do you think a computerized ranking system is superior to the opinion polls?