Preseason Rankings: Do They Matter?

A few days ago, USA Today released its Top 25 Coaches' Poll. You've probably read about it by now, but in case you haven't, click here. The AP Poll, I'm sure, will be coming out shortly, though I don't know precisely when.

And, well, it is what it is. But the relevant question to it all is this: Do preseason rankings really matter? Do they have any real correlation with the actual end result, i.e. the final poll?

Well, that's what I wanted to know.

So, I did some analysis on the subject. All in all, I took the preseason AP Poll from 2002-2006, a five year stretch, and attempted to determine how well the preseason rankings correlated to the final rankings. In doing so, largely for simplification purposes, I decided to only look at the preseason top ten and then look how well those teams did in the final poll.

The following two images represent the entire data set summed up nicely. As you will see, team names are omitted, and only the raw ranking is shown. "Pre" indicates the slot that the particular team was ranked in the preseason AP Poll, and "Final" indicates where the team was ranked in the final AP Poll after the conclusion of the bowl games.

Here goes:





And that is all of the data regarding the subject analyzed here.

So, what did we find?

I ran the correlation numbers, and they were very hit and miss. Some years the preseason poll was pretty accurate, i.e. a pretty strong correlation, and in other years the correlation was almost non-existent. Here is the list of the correlation by year:

2002: .6068
2003: .6973
2004: .7901
2005: -.0105
2006: -.0657

So, as you can see, it's a bit all over the board. Some years you have pretty high correlation, and others there is effectively none whatsoever.

Going deeper than that, once you really look at the data past the raw correlation numbers, you see that even in the best years, the preseason poll isn't too accurate. Though they have a generally high correlation, nevertheless it's still very hit or miss, and the high correlation numbers are mainly due to the fact that a handful of teams were very close in terms of preseason ranking and final ranking.

Let's break things down year-by-year:

2002: As a whole, 2002 was one of the better years, but it still wasn't particularly good. Miami was the preseason number one, coming off of national championship. The Hurricanes didn't really disappoint, finishing second after a controversial overtime loss to Ohio State in the Fiesta Bowl. Aside from that, though, the rest wasn't particularly good. All told, three teams of the preseason top ten did not even finish in the final top 25. Six of the preseason top ten teams did not finish in the top ten; the preseason top ten accurately predicted only four of the final top ten teams. Ohio State, the 2002 national champion that went 13-0, was ranked 13th in the preseason poll.

2003: Much like 2003, 2002 was a good year in terms of pure correlation, but looking deeper it was still very much hit or miss. Three of the preseason top ten teams finished outside of the final top 25, and five of the preseason top ten teams finished outside of the final top 10. LSU, the team that took the Coaches' Poll national championship and finished 2nd in the AP Poll, was ranked 14th in the preseason AP Polll.

2004: The 2004 season had by far the highest correlation of any year, but even it had problems. Though it accurately predicted the eventual national champion (USC), the only time that happened in the five year stretch, and only one of the preseason top 10 teams finished unranked, six of the preseason top 10 teams finished outside of the top 10. The preseason top 10 accurately predicted only four of the final AP top ten teams.

2005: The 2005 preseason poll had effectively no correlation. It accurately predicted the final top two teams, but aside from that it was quite poor. Preseason number three (Tennessee), and preseason number four (Michigan), finished well outside the top 25, and Tennessee went 5-6. Though only two preseason top ten teams finished the season unranked, but five of the preseason top ten teams finished outside the top ten.

2006: The 2006 preseason poll was even worse than 2005. Though none of the preseason top ten teams finished unranked, four teams finished outside of the final top ten. Of the preseason top five, only one team (Ohio State) ranked higher than ninth in the final poll.

At bottom, the preseason rankings can be solid at times, but essentially pointless at others. They generally tell you who the top team is (though, being honest, that's a pretty easy thing to do), but aside from that, I don't think, generally speaking, they really have very much meaning. Every year there are usually two or three preseason top ten teams that finished unranked, and generally speaking around five of the preseason top ten teams finish outside of the final top ten. All told, again, it just doesn't seem like they mean too much. Being brutally honest, it just seems that there is entirely too many uncertainties in college football for preseason rankings to be particularly accurate.

And, well, that's it. A couple of administrative notes:

• I used the AP poll in this analysis, simply because, for whatever reason, it was easier to find the older AP polls than the Coaches' polls. That said, it really shouldn't make a difference one way or the other. All told, the polls are just so similar, there is just no legitimate amount of disparity between the two. They basically come to the same conclusions, and I imagine the analytical conclusions would be the same as well.
  

• For teams that finished outside of the top 25, I counted from 26 on down based on how many votes a particular team received. For teams that did not even receive a vote, I generally assigned them the spot immediately behind the last team that did receive a vote. I understand that there are no official rankings outside of the top 25, but this had to be done to complete the analysis.