Wednesday, October 28, 2009

This Is What You Get For Being Spineless

There are some criticisms of the BCS that I can stomach. "Computers suck!!! Nerdzzz!!! is not one of them. What is most annoying about the criticism is that it is correct in the limited sense that the computer rankings as they are currently used are not very useful. However, this is because the computers cannot take margin of victory into account, so you end up with results like Iowa being #1 in the computers despite a one-point win over Northern Iowa and a three-point win over Arkansas State. Dr. Saturday makes this point adroitly:

Along the same lines, to say "Iowa is No. 1 according to the computers" is a little misleading, since it's more accurate to say that "Iowa is No. 1 according to the computers as manipulated by BCS politics," with the prohibition on margin of victory standing as Exhibit A. Computers can only make judgments based on the information humans choose to give them, and restricting a key piece of information totally changes the results -- for example, both Jeff Sagarin and Kenneth Massey publish their "real" rankings, the set they've developed and honed for years and prefer to use before adjusting the data to meet BCS stipulations, and Iowa isn't No. 1 in either of them. It's not even close, actually, coming at No. 4 in Sagarin's poll (and all the way down at No. 12 in his "Predictor" rating for gamblers) and a humble No. 7 in the estimation of Massey's numbers.

So if you're somewhat baffled by the digital love for the Hawkeyes, don't blame the computers -- blame the feeble human minds behind the system that doesn't trust the machines enough to let them use all the relevant information.

Margin of victory was excluded from the rankings after the 2001 season when Nebraska made the national title game because they did not have a bevy of close wins over overmatched opponents like Oregon did and then the Huskers got hammered by Miami. The subsequent hue and cry about Oregon's omission from the title game ignored the fact that Miami also would have beaten the Ducks like a drum (unless you think that the Ducks' pass defense, which allowed 7.28 yards per attempt, could have covered Andre Johnson and Jeremy Shockey). Based on a sample size of one, BCS critics bitched margin of victory out of the computer rankings altogether, a result that any statistician would tell you is a terrible idea. If margin of victory doesn't matter, then why do gamblers - the people with actual skin in the game as opposed to mere ideological interest - use it?

To put on my amateur history buff hat for a moment, the knee-jerk reaction by the BCS commissioners to castrate the computer rankings is not unlike the various decisions made by the Kennedy and Johnson Administrations that led to the Vietnam War. After the McCarthy era and the incessant "who lost China?" criticisms, Democrats generally and Kennedy and Johnson specifically were terrified of being portrayed as soft on Communism, especially in Asia. As a result, they committed U.S. power to propping up a corrupt regime that could not support itself or command the respect of its people. (This was the same mistake that we made in World War II by ploughing aid to Chiang Kai Shek's inept military leadership, so yay for our policy-makers learning from prior mistakes!) In short, Kennedy and Johnson were a little too responsive to criticism and ended up making a colossal mistake. Does that sound a little like the BCS to you? Anyway, that's what I think about when I look at the quagmire of computer rankings that spit out hard-to-justify results.


mony said...

Actually as I recall, the reason they forced the computers to remove margin of victory (MOV) from their calculation had little to do with Nebraska > Oregon but more with the fact that coaches and media (but not really fans) complained that teams in the BCS hunt were running up the score against overmatched opponents, ruining sportsmanship, etc. Of course, these idiots didn't realize that human pollsters take into account MOV and more importantly, none of the BCS computers rewarded any MOV over 3 touchdowns and in fact, sometimes issued small penalties for huge margins.

Chris said...

Also one thing that is always ignored in "The BCS sucks" articles midseason is that the system is designed to match one and two at the end of the year not during the last week in October

Chip Bankhead said...

Somebody's been reading up Halberstam.

Michael said...

Sadly, I never finished "The Best and the Brightest." Someday.