November 18, 2010
Of the many, many targets in the sights of Yahoo! colleagues Jeff Passan and Dan Wetzel in their definitive anti-BCS manifesto, Death to the BCS, none come in for a more thorough evisceration than the computer polls – mainly from the people who support the nerd element in sports.
Bill James, the patron saint of wonky baseball analysis, issued an open call for serious statisticians to boycott the Series' computers as "gobbledygook math" that uses the veneer of objectivity as a cover to "endorse the coaches' and sportswriters' vote." Before him, a stat professor at UC-Irvine trashed the computers in the Journal of Quantitative Analysis in Sports. One of the algorithms is run by a guy in Oklahoma, Richard Billingsley, who openly admits, "I'm not a mathematician," and doesn't even have a college degree. The rules for the computer polls change too often. They're not transparent. And most of all, they're strictly prohibited from taking into account margin of victory.
Consider that: The models ostensibly designed to compare 120 teams playing 12-game schedules with little overlap aren't allowed to know the scores of the games. A vague notion of "sportsmanship" trumps a legitimately informed process. (Though that certainly didn't stop Wisconsin from running up the score last weekend, or No. 1 Oregon from incinerating opponents by nearly five touchdowns per game, or any other lopsided results routinely taken into account in the human polls.) Even the guys who run the computers can't quite square themselves with that – Jeff Sagarin calls the numbers he submits to the BCS every week the "politically correct" version of his rankings, and Kenneth Massey admits, "it's not the best way to do it."
In fact, Sagarin and Massey both publish rankings every week that represent "the best way to do it," in their estimation, which usually varies fairly widely from the version they pass along to be processed by the BCS. Sometimes, when there are two clear-cut frontrunners (see Alabama and Texas last year), the difference doesn't affect the BCS' stated mission of matching up No. 1 vs. No. 2 in a self-proclaimed championship game – it was going to be Alabama and Texas by any halfway sane measure. The current rankings as we hit the stretch run, on the other hand, are a good example of why they do.
With margin of victory excluded, Auburn is the darling of the "official" polls: Five of the six BCS computers – including the ones run by Sagarin and Massey – ranked the Tigers No. 1 last week for the third week in a row, and the one that didn't (Billingsley) was tossed from their final average as the low score of the six. The Tigers were credited with 100 percent of the computers' first-place votes. Including margin of victory, however, Sagarin and Massey both rank Auburn fourth in their "real" rankings, the ones they would (and once did, before 2004) submit to the BCS if not specifically barred from doing so. Given the broad similarities across the computer polls, it's likely the others (which only publish their BCS rankings) would churn out something similar.
Meanwhile, the potential insurgents take a bath in the translation: Boise State tumbles from No. 2 in Massey's "real" rankings to No. 7 in the BCS, and from seventh to twelfth according to Sagarin; TCU falls from No. 3 in Sagarin's poll to No. 6 in the BCS version, and still finishes third in the official computer average. In other words, without the margin-of-victory ban, it might be Stanford or TCU or Boise State jockeying for position alongside Auburn for the second slot in the BCS title game, instead of waiting in vain for a stumble to open an opportunity to move up. At the very least there would be no question about the Tigers' margin for error in their last two games, because it wouldn't exist.
The point isn't that the "real" rankings provide a more accurate picture than the ones the BCS actually uses (Sagarin's preferred rankings still have two-loss Alabama ahead of one-loss LSU, after Alabama lost to LSU), or that any of the top four or five teams deserve to play for the championship more than the others. It's that their fates are decided in large part by an arbitrary preference that has nothing to do with the reality on the field, and can be completely altered with the addition or deletion of a column in a spreadsheet – again, without changing anything that happened in an actual game.
At the current rate, the difference could cost TCU a shot at the national championship it might have had with a different set of numbers; it could cost Boise State, which is liable to passed over the BCS altogether if it doesn't make the title game, millions of dollars. But when you restrict your raison d'etre to just two teams, those are the kinds of casualties you have to deal with.
- - -
Matt Hinton is on Twitter: Follow him @DrSaturday.