May 12, 2009

Are Preseason Rankings Predictive?

The answer is no. Even if voters want to rank teams in the order of how they think they will finish the next season, they usually do a pretty bad job at it. After my little "unbiased" preseason poll experiment, I started to notice just how bad the 2008 preseason polls look in hindsight. In fact, 2008 may have had the worst preseason rankings in a LONG time.

Does this have anything to do with Georgia being the preseason #1? Not really. Yes, UGA had what many call a "disappointing" season, but the #1 pre-ranked team ended up at #13 and #10 in the AP and Coaches' Polls (respectively) with 3 losses and a bowl win. Sure, this was "disappointing" to UGA fans, but many of those who bleed Red and Black won't be satisfied without an SEC or national title.

So if Georgia wasn't that big of a problem with the preseason rankings, who was? Looking at the AP preseason top 25 and final top 25 for 2008:


Preseason #10 Auburn went 5-7 and ended up (of course) unranked.

Preseason #18 Tennessee went 5-7 and also ended unranked.

Preseason #15 Arizona State went 5-7, ended unranked.

Preseason #9 Clemson went 7-5 regular season, lost bowl game, and unranked.

Preseason #7 LSU went 7-5 regular season and ended unranked.

Preseason #6 Missouri did well with a 9-3 regular season, but dropped to #19.

Preseason #14 Kansas went 7-5 and ended unranked.

Preseason #13 Wisconsin went 7-5, lost bowl game, and ended unranked.

Preseason #19 South Florida went 7-5, won bowl game, ended unranked.

Preseason #20 Illinois went 5-7, ended unranked.

Preseason #23 Wake Forest went 7-5, won bowl game, ended unranked.

And finally, Preseason #25 Pittsburgh went 9-3, lost bowl game, and ended unranked.


By my count, that's 11 teams that started in the preseason top 25 and weren't ranked at all after the bowl games. ELEVEN!

This includes 3 preseason top 10 teams that ended the season unranked (LSU, Auburn, and Clemson). THREE!

Of course this all means that 11 teams that weren't ranked preseason ended up in the final top 25. But guess what? 2 unranked teams in the preseason (Utah and TCU) ended up at #2 and #7. Alabama, ranked #24 by the AP and unranked by the Coaches' Poll in the preseason, of course ended up in the top 10.

So what does this all mean? You can always expect a few teams (especially the lower ranked ones) to end up unranked after starting the season in the top 25. But 11? And also, it has to be unusual for so many highly ranked preseason teams to do so poorly and end unranked. Are the voters just bad at judging who the best teams will be? Or was 2008 just a crazy, unpredictable season for college football?

May 4, 2009

The Search for an Unbiased Preseason Poll

Warning: This is a long, long, analytic college football (and media) post. Read it if you dare. Or, if you're lazy, skip to my preseason polls and comment on what you think about how they look.

Alright, I spent a little while the other day putting together a very early preseason poll for the 2009 season, because yes, the off-season is that boring.

Now first, check out this link here to see an average of a few respected media preseason polls. I'm assuming that, even though there are only a few sources who've released preseason polls, this average poll at the link is very close to what will be the AP/USA Today preseason ranking released later this year. If you aren't convinced, check out the 2008 preseason polls and average and see how they mostly matched up.

Personally, looking at the average of 2009's preseason polls so far, I think there are few things in there that don't look right. Some teams are ranked VERY high (LSU) based on a lot of hype with not a lot of concrete reasoning. Others may be completely ignored in the preseason hype (Missouri) even following successful seasons, based solely on one or two players they lose. (You'll notice that strangely, Missouri, a team that went 10-4 in 2008, isn't on ANY of those preseason polls, all the way through number 41, which includes Vanderbilt and Tennessee).

So I thought it would be interesting to try to make an objective preseason poll and see how it compares.

In my opinion, the most fair and "unhyped" preseason poll isn't a prediction of how you think teams will finish the next season (even though it's fun to predict). Instead, I think a preseason poll should reflect the success of a team in the previous season, with some consideration to how that team will change (or remain the same) in the next.

Here is what I did:

1. I looked at the AP and USA Today final polls from 2008. Then, I found out how many starting players were leaving from each team.

For instance, Florida, ranked #1 in the final poll, is losing only 3 starters from their 2008 team.

2. Then, I added the amount of starters leaving to the rank of the team. Therefore Florida, ranked #1, with 3 players leaving, 3+1=4. Easy, right? I did this for every team ranked from #1 to about #30 in the final polls (using the others receiving votes for the lowerst ranks).

3. I started to make a preseason ranking by starting with the teams with the lowest number. Florida, with "4" as their number, ended up at #1. Then I continued through the rest of the rankings until I reached #25.

Eventually, I ended up with these polls below.

2009 AP Poll (Unbiased)

1. Florida
2. Texas
3. Utah
4. Oklahoma
5. USC
6. Alabama
7. Ohio State
8. TCU
9. Ole Miss
10. Penn State
11. Georgia
12. Virginia Tech
13. Oregon
14. Boise State
15. Oklahoma State
16. Texas Tech
17. Cincinnati
18. Iowa
19. Georgia Tech
20. Florida State
21. Oregon State
22. Michigan State
23. Missouri
24. California
25. Pittsburgh

2009 USA Today Poll (Unbiased)

1. Florida
2. Texas
3. USC
4. Oklahoma
5. Utah
6. Alabama
7. TCU
8. Georgia
9. Penn State
10. Ohio State
11. Virginia Tech
12. Mississippi
13. Oregon
14. Texas Tech
15. Boise State
16. Oklahoma State
17. Cincinnati
18. Iowa
19. Georgia Tech
20. Missouri
21. Oregon State
22. Brigham Young
23. Florida State
24. Michigan State
25. California

Some things to note:

First, these two polls are very similar to the average preseason polls on the link above, with a few exceptions.

-LSU is basically unranked in both polls. While I don't think they necessarily should be unranked, and I do think they could have a great season next year, I consider a top 10 preseason spot is a little bit undeserved after an 8-5 season.

-Some teams that had a lot of success in 2008 are completely left out of the media preseason polls, but show up in the "unbiased" polls. Cincinnati, 2008's Big East champion, is ranked #35 in the average polls, but shows up #17 in mine. Even more baffling is Missouri, which I mentioned before. Missouri isn't even in the top 41 of the media polls. But they rank #23 and #20 in mine.

-A couple of teams that show up in the media polls aren't found at all in the "unbiased" polls. Most notably is Notre Dame, a team that many fans don't want to see in any poll. In the media polls so far, they are ranked #24 after a 7-6 2008 season (including a loss to Syracuse). Another team, North Carolina, is #21 in their polls and unranked in mine.

I also have a few extra statements:

- Yes, Georgia lands at #11 and then #8 in the "unbiased" polls, and averages #15 in the media polls. I didn't purposefully manipulate these things to put them that high. I had no idea how it would turn out until I finished it.

- I know that a few teams don't look right in the "unbiased" polls. Especially the non-BCS teams, like Utah and TCU. These polls might look a little more credible if I put an automatic lowering of rank on each non-BCS team (for instance, dropping every non-BCS team 5 spots). But generally, the media will ignore non-BCS teams in the preseason, and then find them ranked very high in the final poll. There were no non-BCS teams in the preseason top 15 before 2008, but in the final poll, we had #2 Utah, #7 TCU, and #11 Boise State. Whether you think they belong there or not, they will likely end up there.

- In my opinion, these "unbiased" polls fix some of the hype problems of the media. As previously mentioned, I don't think LSU is worthy of a top 10 ranking. I also thought Virginia Tech and Georgia Tech are ranked a little too high in the preseason polls so far (and I'm trying to be objective here). However, I do think Georgia should be ranked somewhere in the 11-20 range, which one of my polls doesn't do. Yes, #8 for Georgia is too high, and I don't think the media has unfairly ranked Georgia so far. And as I also said, some teams like Missouri or even Cincinnati may be dropped a little too much by the media.

Finally, I acknowledge that this formula isn't perfect. I considered weighing the importance of each starter lost and adding that to the polls (IE: losing a star QB would hurt a team more than losing a nationally unknown safety). But it would be a bit difficult to choose the values of individual players, so I didn't bother with that and decided to keep it (relatively) simple. Also, I know that starters returning on a team (and stats in general) often don't mean much in the crazy world of college football. Many very young teams find plenty of success some seasons. But once again, my polls aren't made to predict how the season will END in 2009, but how it will BEGIN, which is what I think preseason polls should be for.

Let me know what you think of them. Do you think the rankings appear mostly fair? Are some teams way too high, or too low? Should non-BCS teams be dropped in rankings just because they aren't in BCS conferences?