A nerd’s-eye view of the BCS

Wired takes a look at the men behind the computers behind the BCS.  It’s not as state-of-the-art as you might think.

… For all the power wielded by BCS computers this time of the year, the machines themselves are hardly extraordinary. In fact, the rankings are processed by individually owned desktop PCs and laptops around the country.

Wes Colley runs his calculations in a database from his home in Alabama. Jeff Sagarin works from his home in southern Indiana, using Fortran, a once popular program used by old-school mathematicians.

Peter Wolfe compiles his rankings baked in C++. Anderson and Hester use a complex spreadsheet and an ordinary HP laptop in Southern California. “When we started, it took Excel half an hour to calculate the rankings,” Anderson says. “Now it takes a fraction of second.”

Of course, these guys think they’re doing the Lord’s work, but Anderson does make a valid point about the new system being somewhat more inclusive to mid-majors than what it replaced.

… Anderson agrees, pointing to the Air Force Academy’s 1971 Sugar Bowl–playing squad, the last team not from a large conference to play in a major bowl before the BCS. “There’s no question in my mind that computer rankings have opened doors for smaller teams,” he says, “Six small-market teams have been invited to BCS bowls in the last six years. It’s only a matter of time until one of these teams wins a championship.”

The weirdest thing in the article is Sagarin’s support for a playoff, not because he thinks it’s a superior way of determining the best team, but because he’s grown jaded.

Sagarin would love to see a playoff, if only for the novelty. “Championship formats are like ice cream,” he says. “I like all ice cream. In that sense, I wouldn’t mind sampling a 16-team playoff, even though I still really like the current flavor.”

It’s as good a reason as some I’ve seen.

About these ads

11 Comments

Filed under BCS/Playoffs, Science Marches Onward

11 responses to “A nerd’s-eye view of the BCS

  1. Turd Ferguson

    I haven’t really made up my mind yet about the whole BCS system, … but I know this: If I had to not only explain, but also justify, this system to someone who was completely unfamiliar with it, I really doubt that I’d be able to do so. And I take myself to be at least somewhat rational and intelligent.

  2. Rick

    As an expert on this sort of thing, I should mention that it is not surprising at all that these calculations can run on a cheap, modern laptop. Even if the ranking formulas are very sophisticated, I would estimate that the average laptop is ten thousand times to a hundred thousand times more powerful than it needs to be.

  3. Macallanlover

    I have personally witnessed human manipulation of a “computer” power poll that is popular in the southeast. Even before I saw that happen, I had serious reservations about computer polls for one critical reason: how can you put the same data in (scores of all games) and come out with widely varying results? (I get that it takes a few weeks to have enough interlocking data/results to link all teams together and give them a power ranking.) When I put a 10 digit number multiplied by any other 10 digit number I get the same result on any calculator in the world. Why doesn’t every computer get the same result for all 120 teams when football results are linked together UNLESS somone assigns subjective values to some of the data?

    • Rick

      There are dozens, if not hundreds, of parameters that could be considered. Each of these could make a weighted contribution to the overall function, and that could be a linear or nonlinear contribution. Although the decisions about what to do with each parameter can be made by an objective process, the selection of that process from among all the candidates that are used on scientific problems is going to be subjective, hence the difference.

      Then there is also ambiguity about what the “best” ranking is. My opinion is that the best ranking is that which best predicts the outcomes of future games (i.e. if team #67 plays team #68, team #67 is more likely to win). This ranking, however, would look bizarre to humans, because you would get weird things like, for example, an unlucky 4-8 team above a lucky 9-3 team.

      • Macallanlover

        I know it isn’t a totally black and white issue, but some of the variations are really pathetic (remember the Seattle paper’s rankings some years?) Agree about the “predictor” method, but you will never get the BCS supporters to get away from simply counting Ws.

  4. Lumpdawg

    This past week’s Jimmy V Classic and ACC-Big 10 match-up in basketball is all I need to know that a playoff is needed. Every year that I watch these preseason tournaments with #1 – #4 match-ups, I wonder why college football doesn’t want the same. Imagine how successful an SEC-Big 10 tournament would be in late August? You could play half of the games over Sat and Sun at the Dome and the other half in Indy or Detroit. UGA is getting $2+MM for the CFA classic next summer, I believe. What would the conference net with a preseason tournament of this ilk? How exciting would it be to play an OSU or Mich to open your season every year without fear that a loss meant your were relegated to Poulan Weedeater Bowl?

  5. Dante

    How complex can the numbers be? You’re dealing with fewer than 200 schools. I’m not surprised those numbers can be crunched on a home PC. And I get the idea that the use of Fortran is supposed to be a knock on Sagarin’s technique but I’d challenge you to tell me what parts of math are different now compared to 20-30 years ago when Fortran was popular. Fortan is an excellent tool for a subset of problems. BCS calculations are very much in that subset.

  6. AthensHomerDawg

    Apollo 13 ….aborted land….. I believe a slide rule was in use somewhere to help get our guys home.

  7. JBJ

    The point is this. We need a centralized system to house the algorithms and the algorithms should be made public. I just can’t believe some of these systems have been devised by people that know absolutely nothing about statistics and probability. We are talking about a huge industry here relying on this craptastic system without any kind of audit on the quality of the data. This is just unbelievable.

  8. Go Dawgs!

    Which tube do you put the votes in?