It's Horn Brain, back with another week of BCS analysis and mortifying (to LA majors) data tables! This column will be morphing through a bit of a hydra state (PB's own lingo) over the remaining weeks as I experiment to find the most accessible form, so go ahead and drop any suggestions (i.e. additional ways to crunch the numbers) in the comments. For now, on to the madness!
OK, so I'm probably not the first to realize this. In my Understanding the BCS story, I came up with a predicting equation for the expected uncertainty in a team's ranking given its position in the BCS. Last week I went through a whole week's analysis without even mentioning it. Well, no more wasted computations, as this week we'll be comparing the uncertainty in a team's ranking to their expected uncertainty, as opposed to the average. What does this mean? On to the madness; That's what it means:
That's kind of a lot of numbers, so we'll look at the deviation (delta) from the expected uncertainty in a graph. 0 is expected, negative is lower and positive is higher than expected uncertainty.
Madness in graphic form:
So who sticks out? Well that would be everyone with a bar above zero, KU, WVU, OU, Texas, Hawaii, Boise, Wisconsin (meh), and South Florida.
Once again we are finding that untraditional and unexpected performances will make computers disagree with the humans. Kansas is in a weird position now, though, since the humans love them (#2 consensus), as well as the computers (#2 avg), yet they still have that abnormal standard deviation. Why? The reason is that the computer top 6 is an absolute gang bang. More madness:
Three teams are tied for third, and I have to think that if the computers would get together and figure out WVU, tOSU, and ASU, then KU would start to lose its grip. I understand that KU won all its games, and that counts a lot this year, but their schedule so far has been absolutely abysmal, and while I completely agree that they're a very good team, I don't think they deserve the #2 spot yet. Beat Mizzou and I'll give it to them, but they seriously have to beat one non-mediocre team first.
The other teams that stick out above the expected values are each the kind of team that you would find as an underdog on the road to a lower-ranked team. Their play has been inconsistent (Texas, UTenn, WVU, OU, and SFU), or they haven't played a tough opponent to really test them yet (Hawaii, Boise). Wisconsin is really just here because they beat an already unimpressive Michigan team that had its legs broken and a mild concussion, so screw them.
These are the teams that give you fits in the Pick 'em game every week, because you never know who will show up (yes, Texas is the perfect example). We'll see over the next few weeks if this holds, but it appears that sticking up above the trendline is a bad, bad, bad sign that you are vulnerable to have your dreams crushed by some nothing Kansas State team, or by some drunken pirate from the desert.
Computers vs. Computers
Maybe you got worried that with all the changes in the column, we'd forget our old friend Billingsley. Let me assure you that we'll have a gay ol' tiiiiiiiime with Billingsley every week, if only because he will never disappoint us.
Analysis: OMG wHy i5 B1LiNgZl3Y 1N Di5 pO1l?!?!??!
Objective conclusion: Billingsley is an idiot, or everyone else is.
Points of note:
1.) Billingsley loves the Big 11. He has tOSU #1 (PB says to riot, people!), Illinois at #11, Wisconsin at #14, and Michigan at #26 (per his site). Maybe Illinois is not that terrible, but his love for the worst conference in the country is totally unfounded. Maybe he has some kind of mechanism in his poll for making sure every conference has three or four ranked teams. Whatever it is, he should be reevaluated after this season, and I mean "reevaluated" in the Franchione/Callahan sense of the word.
2.) Boise State #12, Hawaii #19. If that's not from his totally reasonable stupid reliance on last year's results coupled with his logically sound really stupid limit on mobility from week to week, then he's just writing this poll in an Excel spreadsheet and mailing the printout to the BCS every week in a Hello Kitty envelope with "OFFICIAL" and "IMPORTANT" crayoned all over it.
Lesson of the Week:
I think this week's lesson has a mundane part and a slightly more involved part. The easy conclusion is that you can see who's on shaky ground by comparing their standard deviation with the expected value. Teams that stick out are generally not going to play a complete game, so if they're matched up against someone of similar quality, maybe don't bet the house.
The other conclusion is not really so much a conclusion as it is a hypothesis: Teams performing differently than expected, or that have no real opportunities to prove themselves, will split the computers and the humans, and thus stick out from the expected values. If this is true, it is the first hard evidence I've seen to support the idea that the humans are biased towards the traditional powers. We all have seen a few isolated examples, but if this proves to hold over several weeks, we would have solid evidence.
Hopefully I'll be able to reach a more solid conclusion about this issue after we see the results from the next few weeks. I'll probably do a big wrap up post after all is said and done to flesh that out. Until next week, though, the madness is over, and you can turn the numbers part of your brain off for a few more days.