Wednesday, July 13, 2005

another steaming pile of BCS | by Jay

The BCS finally, formally announced the new poll that replaces the AP. Nothing we hadn't heard already, and still no pollsters named: expect another announcement sometime between now and September 25th. But here's the release. (Interesting tidbits bolded.)

BCS hopes new poll corrects some flaws in system
By Vahe Gregorian Of the Post-Dispatch
Monday, Jul. 11 2005

Caretakers of the Bowl Championship Series unveiled the final details of their annual tweaking and ratcheting Monday. Customarily, they addressed some lingering controversy but were unable to resolve other festering matters - and perhaps again created fresh unintended consequences in the way college football determines its national champion.

In confirming that Harris Interactive would administer a poll that replaces that of The Associated Press, which last December abandoned the BCS out of voter concerns over becoming news makers, BCS coordinator Kevin Weiberg noted that the fledgling poll would not commence until Sept. 25 - several weeks into the season.

The concept of waiting to release a poll speaks to an inherent flaw in preseason voting that Auburn contends damaged its cause last season. Because the Tigers were ranked in the teens in the preseason polls, coach Tommy Tuberville said frequently, they never had a fair chance to surpass preseason favorites Southern California and Oklahoma.

All three finished the regular season undefeated, but USC and Oklahoma met in the BCS title game. Starting later is meant to minimize the impact of preseason guesswork and its logical companion - voting, subconsciously or otherwise, to justify previous votes.

"This will allow for some games to be played during the current season, rather than ranking teams based purely on preseason rankings," said Weiberg, the Big 12 Conference commissioner, who later added, "We have always felt that preseason polls are a weakness of the human polls."

The new poll, to be made up of 114 former coaches, players, administrators and active media members, will be used in conjunction with the coaches' poll and a third equally weighted component (six computer rankings) to form the BCS rankings, the first of which will be released Oct. 17. That essential formula is the same as last season's.

The coaches' poll, however, will be conducted as always from preseason on, Weiberg said, out of the American Football Coaches Association's belief that it serves as "a significant promotion for college football" heading into the season.

Meanwhile, the Harris poll will not alleviate another key point of controversy: Although the roster of voters will be announced before the season, their ballots will be anonymous until the last one of the season - the same as the coaches' poll.

This, despite a clamor for more transparency in the system.

"To make them public throughout the season would mean each week would focus on who voted for whom and perhaps would detract from the games themselves. . . . In fact, there was quite a bit of debate whether or not transparency at the end of the process was even a good idea," Weiberg said, adding, "We will make it available for individual voters to release their ballots as they so choose. There isn't a gag order."

Harris poll representatives said more than 70 percent of its voting panel already had been assembled through a process that began with 300 names nominated by conference offices and Notre Dame athletics director Kevin White. The intention is to ultimately achieve regional balance through 10 names suggested by each of the 11 Division I-A conferences, three by Notre Dame and one from each of the other Division I-A independents.

But Weiberg said it would be incorrect to assume someone nominated by a conference would be considered a representative of that conference. The appeal to conferences for candidates, he said, was merely to generate a cross-section of national candidates, and the idea of having so many voters follows what Harris calls its policy of "best practices and best standards."

Having 100 or more in the sample, Renee Smith of Harris said, adds stability to the results and also has the desirable effect of minimizing the impact of any one person or vote - "potentially a benefit if anyone is worried about biases creeping in," she said during a conference call with Weiberg and the media.

In a statement, Gregory Novak, president of Harris Interactive, said, "We are very pleased that our more than 40 years of polling experience will add independence, objectivity and statistically valid sampling principles to the BCS formula."

And yet this is a new endeavor for Harris, which may or may not be prepared for the wrath of college football nation.

"We're looking forward to the challenge," John Kennedy of Harris said.
Wonderful bromides and platitudes, to be sure, but the problem with the BCS really hasn't been "statistically valid sampling principles" or a lack of "best practices and standards" or "biases creeping in" (okay, well maybe some bias). The only real problem for the BCS has been figuring out who gets to play in the Big Game when there seems to be more than two worthy teams. And this latest mutation does nothing to solve issue Numero Uno.

Jason Kelly had a pretty good rant this morning on this precise point.

If not Auburn, then Oklahoma or USC would have been relegated to a lesser bowl with no hope of a title despite a perfect record.

More than two teams almost always have a case, if not quite as irrefutable as perfection, at least strong enough to cause an annual reconsideration of the procedures.

Leaving somebody out stands as the only identifiable BCS tradition.

In response to this year's issues -- and the AP withdrawing its poll -- the BCS added Harris Interactive and its "statistically valid sampling principles" to its formula.

Good thing, too, because that has been the problem all along. Invalid sampling principles.

How exactly rectifying that issue will help remains a mystery, but the BCS seems to feel most comfortable operating in a fog of data that add up to nothing.

Computer models spit out the best interpretation of facts and figures available to them.

People pretend to identify the difference between California and Texas based on highlights, scoring differentials and how often Lee Corso tells Kirk Herbstreit, "Not so fast, my friend."

From this quagmire of information and speculation, the contenders for the national championship emerge, often too many to accommodate.

So the BCS holds regular meetings, not to debate how to accommodate them all, but how to choose the two that cause the least ruckus.

To hear Weiberg describe their deliberations, they discussed who should vote, when the poll should originate, whether or not to make ballots public, the weight of human impressions versus computer calculations.

Everything but the fundamental issue -- how to winnow the field down to two, and only two, during a regular season that tends not to cooperate.

Even the side benefits of Delay and Transparency that the poll offers in the name of "integrity" are half-assed. September 25th, while later than pre-season, still seems too early if you're really serious about giving the landscape some time to shake out before voting commences. Most teams have only played three games by then, and some teams haven't even dipped into their conference schedules yet. In addition, there are some early high-profile games between big-time programs (OSU-Texas on 9/10, for instance) that are going to result in one of those teams entering the poll with a loss, and potentially allowing a suspect 3-0 team to slide in ahead. (On the other hand, the Coaches Poll will already be going since week one, so who really cares that the Harris is delayed?)

And as far as transparency goes, only the final vote will be on the official public record, so there's still no incentive for the voters to take a long, hard, honest look at things week-to-week. There's still plenty of wiggle room for covert bias and gamesmanship. What's more, instead of professional journalists, who at least presumably have a copy of the SPJ Code of Ethics stashed in their desks, the 114 "former coaches, players, and administrators" have been directly nominated by certain conferences, and probably have only some fuzzy directive from the Harris group to "be honest". Not to say an AP writer can't be ridiculously compromised, but who would you rather have voting on a Notre Dame team: Mitch Albom, or Bo Schembechler? With the former, you at least have a chance.

They just can't get it right.

I mean, seriously. To produce a National Champion, what's better...this:



• 61 Division 1-A coaches
comprising the Board of Coaches vote, counting as one-third, starting at the beginning of the season,

PLUS

• 114 former administrators, coaches, players and current media members, made up of 10 nominees from each of the eleven conferences, plus 3 nominees from Notre Dame, plus 1 nominee from the other major independents, counting as one-third, starting September 25th,

PLUS

• Six computer rankings (Jeff Sagarin, Anderson & Hester, Richard Billingsley, Colley Matrix, Kennth Massey and Peter Wolfe) calculated in inverse points order (25 for No. 1, 24 for No. 2, etc.), and including strength-of-schedule calculations, dropping the best and worst ranking for each team, and adding the remaining four, dividing by 100 (the maximum possible points) to produce a Computer Rankings Percentage, couting as one-third, starting at the beginning of the season.


Or something like this?



(I realize you'd still need some kind of a poll to narrow it down, but at least the poll itself wouldn't be the end of the line, and its impact would be mercifully diminished.)