Thursday, November 10, 2005

Football Wonk: Harris Teetering

I think it's appropriate that the Harris Poll is named for Commandant Harris of Police Academy. Why? Because both suffer public embarrassments, usually involving bruised testicles.

Longtime readers may remember the series on the Harris Poll I did back in August. If you're new to the CHQ, the links are here, here, here and here. My take back then: it's a bad system.

Now that we have the season and games to look at, I thought it'd be interesting to take a look at the current college football landscape and analyze the Harris Poll's system by taking into account the current facts.

1. The Human Flaw Is Still Evident

While not a full-fledged shitstorm, a few bloggers have taken note of Dick Harmon's moronic poll this week. Pat Forde went into detail:

The latest example of a lame system in action comes in the form of a Harris Poll ballot by a Mr. Dick Harmon (2) of the Deseret News in Salt Lake City. Let The Dash say this up front: Harmon has more cojones than the 62 cowardly coaches who vote anonymously in the USA Today poll, or most of the 112 other voters in the Harris Poll. He publishes his ballot every week in his paper. You have to respect that accountability.
... [discusses Harmon's foolish ranking of Penn State]...
The good thing about the Harris Poll is that, with 113 voters, the random lunatic ballot generally is canceled out by the more rational group voting. But there have been so many dumb votes in this stopgap contrivance of a poll the BCS created to fill the void left when the AP pulled out. Every week, something spectacularly stupid pops up in the Harris Poll's small print.
This was the poll that gave us votes for then-winless and still hopeless Idaho. And this is the poll that, this week, included a point for 3-6 Arizona and six points for 4-4 North Carolina State Guys: Get your eyes off the dog-racing agate long enough to take in the records of the teams you're voting for each week. Either that, or put down the ballot and slowly walk away.
Clearly, the Harris Poll's cone of cluelessness extends beyond Harmon. He's going to catch heat this week for being accountable enough to put his bizarre ballot in the paper. The heat truly belongs to the system.

A good take by a consistently good national columnist, in my opinion. Indeed, there are going to be individuals who make mistakes. I admire Harmon for publishing his toilet paper, but it worries me that the other voters are just wiping their asses and hiding rolls and rolls in their closets. I'm not so certain that the random lunatic is cancelled out, since we don't know whether the rest aren't lunatics too.

It is a task of the voters in these polls to have opinions on teams. While opinions are normally rooted in fact, sometimes they aren't. Sometimes the individual possesses not the faculties to transform facts into opinion. Sometimes the individual doesn't care enough to determine the facts at all. Sometimes the individual gets the facts, but is so certain of his own genius that the opinion is divinely inspired yet opposite to the facts. In any event, the voters might not be right. And Harmon's published poll is just one example of this. Now, is it more likely that his ballot is an exception, or the norm?

But that also brings me to my second point...

2. In the Current System, Every Single Ballot Counts. Period.

The view that one or two rogue ballots in the Harris Poll (or the Coaches' poll, for that matter) don't matter in the grand scheme of things is inaccurate. No ballot gets "cancelled out" by other, more rational ballots. This viewpoint is an erroneous view of how the BCS is tabulated.

The "cancelled out" idea relies on the fact that one rogue ballot alone won't drop a team from it's proper position in the poll. If someone hates Texas and leaves them off the ballot, Texas will probably still end up at #2. But it's also possible that some gaps in votes between teams can be affected by a single voter. Example: The gap between Miami and Alabama in this week's Harris Poll was a mere 16 points. If one rogue ballot decided to drop Miami to 20th from 3rd, there's your gap.

And indeed (and this is a crucial point), the BCS doesn't compute its standings from the Harris Poll based on the position of the team. It counts the percentage of points against a total possible. That's why the BCS component from the Harris Poll isn't a simple "1" or "2", it's a percentage number. This is one of the reasons why Alabama is #3 in the BCS, while #4 in the Harris Poll: the gap between them and Miami isn't very great in the actual voting, and thus Harris Poll Component difference isn't large either (.8981 to .8924). Since the total votes (or points) accumulated in the Harris Poll is the actual determining factor, EVERY SINGLE VOTER'S BALLOT COUNTS.

A further example to show this: Penn State was ranked 17th by Dick Harmon. Had Harmon ranked Penn State in line with the rest of the poll (6th), Penn State would have 11 more points, and their Harris Poll component would change, possibly altering the entire BCS standings. Which brings me to...

3. The Butterfly Effect

We all know that a changing a single thing can cause wildly different results. Basic chaos theory. And with the Harris Poll, this is a clear possibility.

Say one (or two or three) rogue voters send in ballots with one particular team dropped without regard to on-field performance or actual facts, like Harmon's Penn State ballot. Not only would the dropped team have a wrongly lowered Harris Poll component, but all the teams slotted ahead of that dropped team would have benefited wrongly and been unfairly elevated. So multiple teams might have wrongly higher Harris Poll components. The result can be a seriously slanted poll, based on just one or two rogue ballots. And that possibility leads to...

4. The Possibility of Impropriety by Voters

I do not presume to know the intentions of voters. I do not wish to accuse anybody of wrongdoing. I would hope that each Harris Poll voter takes his job seriously.

However, because every vote counts, there is a possibility that voters can submit ballots that serve personal or ulterior motives.

In my original posts, I warned of several voters who serve as fundraisers for institutions. Notre Dame stands to gain $14 million by receiving an at-large bid to a BCS bowl. SEC schools' individual cuts would increase by more than $1 million each if two SEC teams receive BCS bids, rather than one. There is a serious monetary benefit.

And where there is a benefit, and an opportunity to influence the bid selections, there is a possibility of impropriety.

Also, I focus on individuals with reason to help their team. Take note that in a system like this, there is just as much motivation for voters (perhaps distantly related to a particular school) to vote against a particular team. Every decision works in two dimensions. For example, If Virginia Tech is close to the #4 slot and an automatic bid, Va Tech and ACC related voters have a reason to vote them higher, but also a motivation to vote the Hokies' closest competitor lower (say, Ohio State). And then Ohio State and Big Ten related voters have the motive to vote not only the Buckeyes higher, but also the Hokies lower. You might suggest that this all cancels out, but remember that not all teams and conferences have the same representation. And it's possible that one voting bloc is more prone to malfeasance than another. Additionally, if it's a close race between Ohio State and VT, who is to say that impropriety with voting for and against both teams results in them both ranked lower than they should, at the benefit of a third team, like Oregon. The point is that the system is prone to this kind of thing.

Below are, under the current landscape of the BCS, situations that could be influenced by the voters in the Harris Poll. I am focusing on the Harris Poll because of the research I've done on it. NB that the Coaches' Poll should be no less scrutinized, as coaches have just as much on the line.

Title Game Selections
Texas, USC, Alabama: Yes, it is very clear that the Tide are probably going to be out, whether right or wrong. However, in the event of a few lackluster performances by the Longhorns or Trojans and huge wins over LSU, Auburn and in the SEC title game, Alabama might close the gap some. In the event that there are three unbeaten teams and somehow the media decide Alabama might deserve a chance, it might come down to voters. Which schools have voters tied to their program?

Alabama: 5 voters (Lacewell, Melick, Townsend, Tuckett and Grace)
USC: 4 voters (Haden, McGee, Munoz, L. Smith)
Texas: 3 voters (Aldridge, Dykes, Mackovic)

Alabama has more voters, but they aren't all that directly tied to the program. Texas probably has as much antagonism as assistance from their related voters.

Also, I assume that no 1-loss team is going to jump over an undefeated team to get a spot in the title game. Sorry, Miami.

At Large Selections
Assuming that Texas, USC, Alabama, Miami, Penn State and West Virginia receive an automatic bid for winning the conference, the following schools are jockeying for the other two at large bids: Virginia Tech, Ohio State, Georgia, LSU, Texas Tech, Oregon, Notre Dame. These teams are all in the Top 12 of the BCS standings right now.

In order to qualify to receive a BCS at large bid, a team must have 9 regular season wins and finish in the top 12 of the BCS standings. That is the minimum. But for merely qualifying leaves the choice in the hands of the BCS bowls. The BCS rules do provide for automatic qualification, which leaves the bowls' choices out of the equation. I will focus on that.

To automatically receive an at-large bid, there are three methods: (a) for teams outside the Big 10, Big XII, ACC, Big East, SEC, Pac-10 and Notre Dame, the team must finish in the top 6 of the BCS standings; (b) For teams in the Big 10, Big XII, ACC, Big East, SEC and Pac-10, the team must finish in the top 4 of the BCS standings; (c) For Notre Dame, it must finish in the top 10 or have won 9 games*

*: I believe this is an unclear rule - the BCS website suggests that Notre Dame is automatically qualified for a BCS at-large selection "If one or more teams other than Notre Dame qualify for automatic selection under this provision [addressing non-BCS conference teams ranking in the top 6], Notre Dame shall ALSO qualify provided it is ranked in the top 10 or has won at least nine games." To me, that reads that Notre Dame only qualifies AUTOMATICALLY if another non-BCS conference school also qualifies, which under this year's factual scenario, wouldn't take place. By the plain terms of the BCS site, it appears that Notre Dame too would have to finish in the top 6 of the standings, since they are an independent school. The "top 10/9 wins" language only appears after the modifier where another non-BCS team has qualified. I think that in practice, Notre Dame will be the first team selected if they are in the top 12 and there are not 2 automatic qualifiers forced to be selected ahead of them. Also, starting next year, Notre Dame must finish in the Top 6 with 9 wins to receive an automatic bid (and their $$$ share drops). That clears things up.

As things currently exist, those 7 teams are jockeying for positions 3 or 4, assuming that losses by teams ahead of them don't cause them to win the conference and obtain an automatic selection. Let's look at this team by team.

Virginia Tech: Highest ranked team not currently in position to win the conference. They sit 6th. Could move into 4th by winning out and would still not win the conference. Strong in computer rankings, but need to improve in the polls, which is where human tinkering might play a role. Harris Poll voters related: 1 (Bill Dooley). Not much chance for help directly. The conference has 18 voters related to the school though, and they all have an interest in having 2 ACC teams in the BCS (not just in BCS revenues, but also in the way schools with worse records move up in the Bowl bid ladder - i.e. Georgia Tech goes to the Peach Bowl instead of the Meineke Bowl).

LSU: Next highest ranked team, but would win conference if it won out. A loss this weekend forces them to fight for only an at-large slot, but their standing now would change dramatically. Very high in the polls now, need help in the computer polls. Something to watch for is how far they would drop if they lose. Harris Poll voters related: 4 (T. Lewis, Moody, Stokely, Townsend). Also the SEC has the most related voters (27).

Ohio State: Similar to VT, in that they can win out and not win the conference. Also, their high computer average and room to improve in the human polls provides an opportunity for individual voters. Harris Poll voters related: 4 (Bruce, Geiger, Jacoby, Roda). 21 voters have ties to the Big Ten, which would benefit from having 2 teams in the BCS. Andy Geiger's recent ties to the university should be noted.

Georgia: Similar to LSU, in that if they win out, they would earn an automatic bid by virtue of winning the conference. Losing another game probably would prevent them from earning an at-large bid. Harris Poll voters related: 2 (Goff, Bestwick). Again, the SEC has the most related voters.

Oregon: Similar to OSU and VT. Room to improve in the human polls. Can possibly get close to the top 4 if it wins out. Harris Poll voters related: None. The Pac 10 has 17 conflicts. If there is going to be malfeasance, Oregon probably won't benefit. The cards are definitely stacked against them. Also, the fact that their poll rankings are lower than their computer rankings, combined with the lack of an Oregon voter in the Harris Poll should call into question the integrity of the system.

Notre Dame: As stated above, I feel the rules governing Notre Dame are unclear. I'm uncertain whether Notre Dame would receive an automatic bid if they were ranked in the top 12, 10 or 6. Also, whether the rules require them to be ranked there in addition to having 9 wins, or whether 9 wins alone would be sufficient (the BCS site uses the word "or" to suggest the latter). Also, it is unclear from the BCS website as to the priority of Notre Dame when compared to a top 6 non-BCS conference team and compared to a top 4 BCS conference team. In any event, it probably doesn't matter whether Notre Dame needs to get to 6th place, 10th place or 12th place to get an automatic bid. Under the factual scenario as exists today, there will probably not be two BCS conference teams who didn't win their conference at #3 and #4, so likely there will be a "non-automatic" selection to be made by the BCS Bowls. And it is likely that Notre Dame will be that team. So in practice, the number they need to look for is #12, and they are already there. Additional security can be found with the fact that Notre Dame has 5 related voters (Haden, Ismail, McConnell, Morse and Valdiserri). Remember that Morse is a "major benefactor" to their program.

Texas Tech: A longshot to get in the top 4, but it is possible they could win the rest of their games and have several teams ahead of them currently lose. Their computer ranking is currently higher than their human poll ranking as well. Harris Poll related voters: 3 (Dykes, Holub and Lawless). Remember that Holub is a current fundraiser for their athletic program. The Big XII has 23 related voters, second most to the SEC.

And just so I don't leave anyone out, TCU (17th) and Fresno State (22nd) both have a long way to go to compete for a Top 6 BCS rating. Fresno State can make a big statement with an upset of USC, but it probably still wouldn't be enough. Either of these teams have to get to #6 in the BCS standings to qualify for an automatic bid. TCU has one related voter (Windegger) and the Mountain West has only 9 related voters (second fewest). Fresno State has three related voters (Lengyel, F. Lewis and Sweeney) and the WAC has 11 related voters. Sweeney is a current fundraiser for Fresno State.

So that's the situation as it is. I'm on record as saying the Harris Poll is a problem and calls into question the integrity of the game. Whether or not the situation this year will be close enough to have easily determinable malfeasance is still in doubt. Even if the situation doesn't present itself for there to be vote rigging (or trading, or any other problem), the fact that fraud is even possible should be reason enough to reform the system.

Further, I believe college football relies on a narrative as the season goes along. Listen to the way GameDay pundits use the phrase "how this is playing out". Read columnists use similar terminology. Just because the final vote isn't close doesn't mean that a few votes here or there in the earlier Harris Polls might not have changed the way teams were covered. One simple erroneous vote can slot a team a little bit higher, which in turn garners them more attention from the national media, which voters rely upon to vote in the next week's polls, which affects the next week's coverage, etc.

Unfortunately, chaos theory appears to be alive and well in college football, with biased individuals flapping their butterfly winged Harris Poll submissions.