Poll Dancing: Week Thirteen, or Do You See What I See?

My investigation into the secret bias of the BCS computers continues with something of a revelation this week.  As the season winds down with this weekend’s conference championship games (as well as offerings from non-championship conferences), the BCS computers are grinding at full speed in their effort to boost the position of the Big 12 (and, to a lesser extent, the SEC, which doesn’t really need any help).

I linked this article in yesterday’s post for a different reason, but it’s worth pointing out Wetzel’s main complaint: that the computer formulas used to generate the standings are kept almost completely secret (only 1 of the 6 is even available to the BCS itself).  He wonders (legitimately) why we should trust that there isn’t something shady at play, and I have to agree.  It’s simple really: if there’s nothing to hide, why all the hiding?

Having compared the average ranking of each team in the human portion of the BCS standings to their ranking in the computer portion, I think I have an idea what might be up.  You might recall that the reason I started monitoring this in the first place was what I felt was an abnormally high ranking for Oklahoma State a few weeks ago.  As the digging continued, it became apparent that the entire Big 12 was being favored more heavily than any other conference in the computers.  This week, I noticed something else when examining the numbers, so I sorted them in order from most “overvalued” to most “undervalued.”  (These are generic terms to illustrate the difference in the rankings, they are not intended as an assessment of a team’s ability.)  Here are the seven most undervalued teams–the teams that receive significantly worse (at least 2 full spots) rankings in the computers than in the human polls:

Houston (-2)
Oregon (-3)
Southern Miss (-6)
Michigan State (-6)
Wisconsin (-6.5)
Virginia Tech (-6.5)
West Virginia (-8.5)

In this group, we have the probable Big East champion, the probable ACC champion, the Big Ten champion, the probable Pac-12 champion, and the Conference USA champion.  Boise State and TCU, one of whom will win the Mountain West, are undervalued by one spot each.

Every champion (or probable champion) of every automatic-qualifying conference besides the Big 12 and SEC is undervalued by at least three spots.

The contenders for the SEC crown don’t fare that much better, Georgia is undervalued by one spot while LSU breaks even, ranking #1 in all polls by virtue of being undefeated.  By comparison, probable Big 12 champ Oklahoma is overvalued by 5.5 spots, while possible champ Oklahoma State is overvalued by 2 spots.

Is the Big 12 really that much better than every other conference?  Are the computers programmed to treat Big 12 teams more favorably?  Are there even any computers at all?

Comments

  1. So not being able to see the formulas obviously leads to speculation, and this answer to your question will become circular and probably collapse on itself, but the overvaluing of the SEC and B12 point to the computers “figuring” that those conferences (and the subsequent opponents) are stronger. Which becomes a self-fulfilling prophecy.

  2. An issue I see with this analysis is that you are using one biased system and comparing it to another.

    In a totally unbiased system that awards a single point for each FBS team beaten, and another point for each of the schools that they beat, and a point for each of the schools they beat to 10 tiers, Oklahoma State is by far the most accomplished team. Simply they beat the most teams who beat the most teams. Here is my top 10 Ranking:

    Rank School Pts Wins
    1 Oklahoma State 422,675 11
    2 L-S-U 313,984 12
    3 Oklahoma 308,202 9
    4 U-S-C 300,743 10
    5 Kansas State 297,569 9
    6 Oregon 289,185 10
    7 Stanford 286,927 11
    8 Baylor 262,677 8
    9 Virginia Tech 231,301 10
    10 Clemson 199,251 9

    Anything else you add to this could be considered bias. Home vs. Away, margin of victory, etc. All of those things bias one “kind of play” over another and everyone arguing for anything other than wins on wins will bias toward something their own school or conference does better than another.

    Even my ranking biases for FBS wins, because I am too lazy to try to figure tiered wins for other divisions, as well as I am biased and say that beating a div 2 school isn’t really a win, so I award no points.

    However to answer the questions from the post: 1. There is everything to hide, millions of $$$ are traded on the outcome of those formulas. 2. Big 12 is not that much better, however they tend to play fewer Div 2 games on average than other conferences, In fact the whole conference only played 6 FCS games, compared to the SEC/12, B1G/10, Big E/8, ACC/8, so that should count for something. 3. Programs probably do not overvalue the Big 12 more than either the B1G or SEC. 4. There probably are not computers, and if there are, the final results probably get altered to fit some preconceived bias.

    I know from my first year running a computer poll, you tweak things and refine your algorithm so that your outcome “makes sense”. I had to stop doing it because you start to notice how bad the bias gets.

    Notice how Alabama does not even show up on my top 10, I know, it “looks” weird, but if you dig into their wins, they are not that impressive unless you include significant SEC bias as Arkansas was their only quality win all year, but all you hear on Sports Center is 3 hours of droning on about their single quality loss. That is bias.

Leave a Reply to el Kaiser Cancel reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: