Dr. Cinderella or: How I Learned to Stop Worrying and Love (?) the BCS

In the entire known world of metrics, nothing is as important as the American college football rankings.

But if you believe the BCS rankings, then you probably believe in yesterday's weather reports, and you're probably not a gambler. Or, you believe some teams go to Hell, while other teams go somewhere else.

For the most part, it's really a given that most teams will go to hell at some point, and it is well known that when ranked teams go to hell, they go to a harsher hell than do most unranked teams.

All teams try their best to go to hell in the preseason, under less scrutiny. Whether it's preseason or not, good teams may go to hell only for a brief stretch of a practice, or in season perhaps during one quarter (probably the first or the third) of an actual game or two. You can see it on the coach's face: "Well, that just went all to hell."

But back to the BCS and metrics. If your team is ranked say number seven in the top twenty-five in the BCS list, does that mean that the teams ranked higher than your team are "better teams", and the ones ranked lower are "worse"?

Of course not. At best, it means that higher-ranked teams had a higher frequency of important beneficial results from the circumstances they met while operating. The key part of that explanation is the slippery part: "important".

With the formula used to calculate the BCS rankings, there is not nearly so much debate about what factors are important to each team, as there is debate about whether the factors important to Team X are also equally important to Team Y. In a measurement system like the BCS rankings, the point of the formula used is to impose it as an identical calculation on all teams. Whereas in gambling, the beauty of its complex logic is that determining the probable winner of a contest emphasizes what each contestant uniquely needs to have the best chance to win.

Another very interesting counterpoint to the BCS rankings would be Wall Street. Should we think of the BCS ranking as being analogous to the "stock price" of a team?

Well, even given the similar degree of analysis that goes into each contributing poll underlying the total BCS calculation, stock price is based on much the same idea that gambling is: given some degree of uncertainty, what is the probability that the contestant will actually generate the critical advantage it needs to have its unique formula for success prevail in its expected future circumstances? Unless contributing polls focus on that, there's not much reason to "put stock in" the composite ranking.

But so what, we still want to use the BCS ranking as an indicator of how probable Team X is to defeat Team Y in a hypothetical matchup. There, our desire reflects our interest in relying on "intelligence" -- on the power of facts to reveal truth (or at least repeatable or persistent patterns) -- just not nearly so much as we pretend it does. The determinant feature of this ambition is the matter of whose intelligence we choose to use, because we rigorously choose to not use other intelligence outside of that. And as we know from "experience", the flaw in this approach is that probable outcomes in sports are determined as much by what we don't know as they are by what we do know. And guess what, this limit on the BCS exposes that the real point of the BCS must be something else -- namely, not to find the best team but merely to promote an agenda of creating artificial interest ("need") in order to then make money by satisfying it.

On that note, it becomes apparent that an ideal, non-commercial performance ranking system (not the BCS) would not be a weekly list of mainly historical accuracy comparing apples and oranges to each other. Instead, it should be a monitoring system that takes the current probability of a team's next future win, and describes the importance of that probability relative to the importance of each other team's similarly described outlook.

Let's say that a largely uncelebrated team from a so-called "minor" league has a high probability of winning all of its games, and so does a high-profile team from a so-called major league. What is the relative importance of the minor team's perfection versus the importance of the major's? The real importance is not that we think the one team would probably beat the other in a playoff.

Instead, the real importance is in how each team's own league (or said differently, its market) can respond to the performance of the respective teams. The major league team can't do very much (that it likes) with the excellence of the minor league's team -- or at least not much more than the major league's scouts and recruiters have already done. But is there some important relationship between the minor league team's excellence and the major league team's league? How does the minor league team benefit (i.e., "play in") the major league market? Even the major league as a whole is lukewarm to the good minor league team. The minor league, however, can of course do a lot with its team's excellence.

The point would be that even if the minor team is in the BCS top ten along with the major team, but they don't actually play each other, it doesn't matter and it doesn't mean that the higher ranked team (whether major or minor) is the better team.

With that admission, it's easier to relax and understand what we do want the BCS for: we want it to suggest who should go to the playoffs to make the playoffs more interesting -- at which point the bookies (stockbrokers, traders, stakeholders) can take over and do their job.

Oh... wait... There are no playoffs...

Soooooo...

What's wrong with the BCS? Nothing, really, except that it's a gun pointing at a non-existent target.

Is your own measurement system loaded but pointless?

[Pick that Nit Dept. -- at least the BCS, or "Bowl Championship Series", isn't actually called the "National Championship Playoff". A series of bowl champions we can live with; why not? All ya gotta do is like bowls.]

Posted by Malcolm Ryder at 7:02 AM | Comments (0) | TrackBack

October 6, 2007