Stat Geek Perspective: Preseason College Football Polls

August 9, 2011 - by David Hess

The preseason USA Today NCAA football coaches poll was released late last week. Oklahoma nabbed the top spot with 42 first place votes, followed by Alabama (13), Oregon (2) and LSU (2). But how important is this initial rating? What does it tell us about how a team’s final rankings? How would we have voted, if we had a ballot?

While the final pre-bowl polls are vitally important in college football, and are undoubtedly influenced by the initial ordering (if several teams go undefeated, the ones that begin the season ranked higher will likely finish ahead as well), these preseason ratings are not gospel. To see how accurate the pollsters tend to be, we took all of the preseason polls from the BCS era (1998 through 2010), and looked at how well they predicted a team’s final ranking.

Average Final Ranking

Let’s start simple. It turns out that being voted one of the top couple teams is a great indicator of success, but the value of a high ranking drops off sharply after that. Given a team’s preseason ranking, and no other information, here’s a chart that shows the best guess for their final rankings. From the raw data, it looks like voters do better at grouping teams into tiers than they do at ranking teams within each tier. There are several distinct bands: top 2; #3 through #8; #9 through #14; and #15 through #25.

Avoiding Disappointment

Since 1998, no team ranked #1 or #2 in the preseason has finished the year unranked, but nearly one fourth of the teams ranked #3 or #4 in the preseason have dropped out of the polls. For those teams that barely sneak into the first poll, the outlook is grim — roughly 75% have ended the year unranked. Here’s a chart showing the odds of finishing the year unranked, based on a team’s preseason ranking:

chart plotting fraction of teams that finish unranked, versus preseason ranking

The Upper Echelon

Now let’s look at the chance of a positive outcome. Again, being in the top two has historically been a great sign, as 75% of teams with preseason rankings of #1 or #2 have landed in the top 5 at season’s end. No other ranking slot has even a 40% success rate by the same metric. Last year’s Auburn team was an aberration, as teams ranked worse than #20 have historically managed to crack the final top 5 only 3% of the time:

Our Take On The 2011 Preseason Poll

While the preseason ratings used to create our projected standings aren’t currently listed on our site, we can dig into our database and pull them out for comparison. Doing that gives us a few key take aways:

  1. We agree that Oklahoma should be the best team in the land. The Sooners are our #1, and #1 in the Coaches poll, and it’s not particularly close in either instance.
  2. We would slot Oregon over Alabama for #2, at least if we’re arguing over who is the better team. That said, it’s fairly close, and our ratings actually show Oregon with a tougher schedule, which leads us to the next point…
  3. Oregon and Stanford were highly rated, but the rest of the Pac-12 deserved some love as well. Cal, USC, Utah, Arizona, and Arizona State are all in roughly the #15 to #30 range in our preseason ratings. Even if we’re overrating the entire conference, at least one or two of that group ought to have snuck into the top 25.
  4. The wrong Florida team is in the top 10. We rate the Gators as better than the Seminoles this year, and view Florida State’s  #5 ranking as a reach.