10/26/2010

machines smarter than humans, again

I was prepared for this to be a slow, sad sports season, really. The Chiefs and Rams are coming off of horrific stretches. The Big 12 is saying farewell to Nebraska, heading east, and Colorado, looking west. And there are More Important Things, from the economy and midterm elections to new phases for family and friends. Not to mention that I'm behind on computer games.

But dang nabbit, something is in the water this fall. After a hot summer, this fall has been fantastic. Missouri's NFL franchises are 7-6. That's a winning record; better seasons for both teams have already been assured and we're less than half way through this season!

The Missouri Tigers have also won 7 games. Without losing any. To top it off, the Spartans are undefeated as well, while the media darlings of Alabama and Ohio State have both lost a game already.

And that's where this all comes to a head. The silly season of incoherent poll voters is in full swing. For all the incessant yacking and yammering about how awful the BCS is, the main problem is that the human polls account for twice as much weighting as the computers. (human) Pollsters that do things like looking at prior years instead of the current year. Or assign present values based upon what they believe will happen in the future instead of what has already happened up to this point. Or apply different criteria to different teams. It's truly bizarre, and this season demonstrates the madness in several ways, particularly looking at the most basic things in sports: the win-loss record and strength of competition.

In years when there is lots of chaos caused by few or no teams with unblemished records, the challenge becomes how do you rank teams that have lost games. But the first half of this season is comprised of something rather straightforward: There's an undefeated team in the Big 12, the SEC, the Big Ten, the Pac 10, the Mountain West (which has two), and the WAC. It's pretty simple. This is the absence of chaos; it's called order. These are the teams you talk about first, before looking at the bigger picture. And when you start actually looking at 2010's undefeated teams, I'd argue it becomes pretty clear that there are four teams that have distanced themselves from the pack. None of them, by the way, have the names of Alabama, Ohio State, Oregon, or Boise State. Nope, the top tier of teams for the first half of the 2010 season are, in alphabetical order, Auburn, Michigan State, Missouri, and TCU.

In order to rank a team above those four, some other consideration has to come into play. And in order for that to be legitimate, that consideration has to be applied to all teams. That's the key point: consistency. The computers bring it in spades. The polls, well, sometimes I'm not sure what they're smokin'.

Here's how the BCS computers rank the top four teams:

1. Auburn
2. Missouri
3. Michigan State
4. TCU

Here's the Harris and USA Today polls:

1. Oregon
2. Boise State
3. Auburn
4. TCU

To demonstrate the overwhelming stranglehold of the humans over the BCS - even though all you'll hear in the media is pundits decrying the horrible choices of the computers - is to look at the combined rankings. Oregon and Boise State remain in the top 4.

And this leads us to the natural question: what consideration is being applied equally to reach this kind of disparity by teh humanz? The computers think the SEC leader should play the Big 12 leader for the national championship, with the Big Ten and Mountain West leaders being the runners up. The human polls don't put either Auburn or Missouri in the top 2, and they don't include Missouri or Michigan State in the top 4 at all!

Poll voters aren't just not biased against Boise State and Oregon. They're rabidly biased in favor of them, to the point of crowding out the real story lines of the first half of the season. In fact, there's such a drop-off after my top four teams, that the computers handle this in an interesting way. They put one-loss Oklahoma at number 5. And one-loss LSU ties Boise State for sixth, then Oregon at number 8.

Then there's the matter of one-loss teams. There's no way around the human stupidity of ranking Alabama so highly. None. When Missouri loses to Nebraska, then you can rank Alabama higher - but that loss hasn't happened yet! It's not just Missouri that's slighted by Alabama bias, though. Every one loss team suffers. You see, not only is Alabama not on some semi-pro uberwonderseason, but they haven't even put together the best first half of the season among one loss teams. Their loss at South Carolina is worse than Oklahoma's loss at Missouri, or Wisconsin's at Michigan State, or Ohio State's at Wisconsin, or Stanford's at Oregon, or Florida State's at Oklahoma, or LSU's at Auburn, or Oklahoma State's loss to Nebraska. In fact, what's remarkable about Alabama's loss compared to other top one-loss teams is how badly they got beat by a lesser team than the victors of all those other one-loss teams.

Oh, but it's the quality wins you say, not the losses? Ah, interesting. What teams have Oregon and Boise State beat again (this year) to warrant placing them ahead of Auburn, Missouri, Michigan State, and TCU? Uh-huh. And then when we look specifically at Alabama, they haven't beaten anybody with fewer than two losses. Wisconsin has beaten Ohio State. Nebraska has beaten Oklahoma State. Oklahoma has beaten Florida State. Heck, even Stanford's win over USC is as high a quality as any victory on Alabama's schedule so far.

The computers recognize this, too, and rank Alabama accordingly: at number 12. Alabama is a great team, one of the best in the country. They just don't belong in the Top Ten until they beat Auburn and LSU - which, as of now, they haven't done - and until Oklahoma, Missouri, Nebraska, Utah, Wisconsin, and Ohio State lose more games - which, as of now, hasn't happened, either.

So if it's not being undefeated, and it's not the quality of wins, and it's not the quality of losses, how are pollsters making decisions? Is it the conference?

Well, TCU is in the only conference with two unbeaten teams, and other schools like Air Force and San Diego State are playing competitively, too. That's all the more reason to rank them ahead of Boise State. The Big 12, meanwhile, is putting on a conference finale for the ages. Oklahoma, Missouri, Nebraska, and Oklahoma State are all quality programs. In the Embarrassment of the Season Bowl, Colorado beat Georgia. Heck, Baylor is bowl eligible, Texas is simply middle of the pack, and Kansas managed to beat Georgia Tech. In the Big Ten, meanwhile, Michigan State, Ohio State, Wisconsin, and Iowa all look very solid. Oh yeah, Michigan has a guy who can make headlines. It's also interesting that Iowa is ranked so much more highly by humans than machines; why wouldn't the Big Ten love be going to Michigan State, first and foremost? And Illinois beat the same Penn State team Alabama did. In a bizarre twist, the SEC luster isn't rubbing off on the conference's undefeated team, or even the team it beat. Rather, it's the third best SEC performance of the first half, Alabama, hogging all the (human) headlines.

After all, we can't have preseason prognosticators bothered by how the season actually turns out. I mean, imagine if the World Series wasn't the Yanks and Phillies? How could Texas and San Francisco even be worth talking about?

Unfortunately, this post has to be this long, because it's absolutely necessary to point out the inherent contradiction which the humans can't shake: it simply can't be that Boise State and Oregon are having great years while it is also true that Alabama is having a great year. The second half of the season could very well play out differently, particularly with possible losses for Missouri at Nebraska, Michigan State at Iowa, Auburn at Alabama, and TCU/Utah. But as we mark the roughly halfway point through the season, it needs reminding in all the biased pundit hype that so far in the 2010 season, the computers are smarter than the pollsters. The problem with using coaches to rank these things is that coaches have a far more important demand on their time - coaching. There's no evil conspiracy here, just an unfortunate problem with relying so heavily on the human polls. Fortunately, there's a simple solution. Have the computer rankings account for half the score, rather than 1/3.

So, here's where I place things. I will re-arrange accordingly after my higher ranked teams actually lose and the lower ranked teams actually win. And remember, the computers will adjust, too. They have no pre-season prediction egos to feed. That's probably why the media blathers on about how horrible the machines are. Change is scary.

1. Auburn
2. Missouri
3. Michigan State
4. TCU
5. Oregon
6. Boise State
7. Utah
8. Oklahoma
9. Wisconsin
10. Ohio State
11. LSU
12. Alabama
13. Nebraska
14. Florida State
15. Oklahoma State
16. Stanford
17. Arizona
18. Mississippi State
19. Iowa
20. Baylor
21. South Carolina
22. Arkansas
23. Virginia Tech
24. Miami
25. San Diego State

No comments: