Monday, September 28, 2009

The Fatal Flaw of the BCS

What's wrong with the BCS? Look no further than the two polls that account for two-thirds of the BCS Standings.

This week's Simulated BCS Standings are just about as good as the real thing, with nearly 90 percent of the data for the actual standings available - both the Coaches and Harris polls and four of the six computers.

But the near-fatal flaw of the BCS Standings becomes immediately evident as soon as the Harris Poll is released. With another year potentially to be dominated by parity - only 17 teams remain unbeaten after four weeks - the standings will have the final say on who gets to play in the BCS national championship game. And yet again, the standings will be heavily influenced by two highly biased and unreliable polls.

Let's count the ways:

1. The Harris Poll doesn't come out until the fourth week of the season, ostensibly to free the voters from preseason prejudices. Guess what? It's just not happening. The top 10 of the Harris Poll has the exact same 10 teams from the AP Poll, and in the exact order, except for Nos. 8-9 Ohio State and Oklahoma traded places.

2. If the Harris Poll was to be free of preseason biases, how would you explain the fact that Oregon is ranked No. 23 and Cal at No. 21? The Ducks just destroyed the Bears by 39 points. The two teams have identical records at 3-1. And Oregon played a considerably tougher schedule than the Bears did, with games against three ranked opponents. If the Harris voters indeed voted on the basis of pure performance, there is no way that Cal should be ranked ahead of Oregon. ... The same goes for Penn State being two spots ahead of Iowa.

3. The credibility of the Harris Poll is again highly questionable, with voters putting teams like Iowa State, Kansas State, Washington and Texas Tech in the top 25.

4. The Coaches Poll has the same sort of problems, as usual. And since both polls do not release the individual ballots to the public until the final vote, there is little or no transparency in the process.

As long as these two polls get to dominate the BCS Standings, the BCS will always lack a certain amount of legitimacy. With so much money and prestige on the line in the biggest intercollegiate sport, it's in many ways mind-boggling that a system can continue to thrive with so little checks and balances to ensure fairness.

Then again, big-time college sports isn't really about fairness, is it?


Brian said...

I calculated the BCS Standings under the old formula where human polls and computer polls were equal and strength of schedule was an explicit part of the formula. Here are the top 5:

1. Florida, 2. Alabama, 3. LSU, 4. Texas, 5. Boise State

(Using this formula last week, Florida was 9th.)

Just thought this would be interesting to note.

Anonymous said...

And yet how is it fair that two computers have LSU ranked in their top 5? LSU should have lost to Mississippi State! Ok, they won, but how perverse that that kind of winning gives a boost both in the computers and with voters while tOSU losing a close one to USC is a pretty big negative in both elements.

At least there is a majority of poll voters that can take context into account. The system correctly gives one voter little overall impact in the process while one computer can have a huge impact. Ok, I understand that the two outliers are thrown out on the computer polls, but their validity isn't any less credible the the entirety of the polls.

Keith said...

So people hate preseason polls, yet the Harris doesn't prove to be any more of an advantage. Must be the human element, right?

OK, so computers can eliminate the bias (we're told), but people hate them too because early on there isn't enough data to rank Florida, et al., over ULa-La.

So what happens? The fans and media complain then show their bias with their own Top 5, 10, etc.

If anyone catches themselves or someone else arguing about which team is better -- especially in late November -- and those teams didn't/won't play each other, then they're making an argument for a playoff whether they like it or not.

Until that happens, in whatever format, the mind games of the 'beauty pageant' system that we have now will replace the real games that could be taking place.

In either situation there's going to be arguing. So why not have a system where the arguing is not over Top 5 teams, and actually meaningful, put-up-or-shut-up games are played with matchups that might otherwise never happen?


LAprGuy said...

Meanwhile, the computers use non-conference records to help validate/rank the conferences, affecting each teams' overall score indecices.

Oh, look, the SEC is 23-2 out of conference -- Gee, their teams must be the best.

dethwing said...

LA: Those conference rankings are totally unrelated to the team ranking. It's simply applying his overall method to the conference games.

LAprGuy said...

Dethwing - Your specific point about Colley is correct.

However, my point that "the computers use non-conference records to help validate/rank the conferences, affecting each teams' overall score indices" stands: The conference's non-conference W-L is a stated factor in some of the computer rankings. So going 23-2 -- by playing virtually no games against other BCS conferences -- looks good in some of the formulas.

But I hear ya' and concur.

I think the fatal flaw of the BCS is trusting mathematicians to "rank" teams. Amusing: Colley's thesis (I read it) expresses surprise -- and, therefore, validation -- that just by crunching numbers, he is able to virtually match the tops of the human polls.

REALLY, Colley?! You don't think it's because going 12-0 or 11-1 is going to be judged equally good by humans or by calculators?!

(Of course it helps that the humans also overrate the SEC's 23-2 cumulative record ...)