By Stewart Mandel, SI.com

Ted Ginn Jr. and Ohio State ran past Michigan to solidify their No. 1 ranking last November. Bob Rosato/SI

So as long as there are going to be preseason polls, one might as well gain an understanding of how they work. I've been studying and conducting these things myself for the better part of a decade and in doing so have noticed four common criteria most voters emphasize:

1) Number of returning starters: You hear this statistic a lot in the months leading up to the season. Because it's nearly impossible to gain an intimate knowledge of 25 different teams' entire 80- or 90-player rosters, we focus primarily on the 22 starters. A team that was pretty good the year before and returns 16 or more is generally considered to be loaded. Ten to 15 returning starters is about average and a sign of a fairly experienced team. Single-digits is considered a rebuilding year. While this statistic is a decent starting point in determining the strength of a team's roster, it hardly tells the entire story. For one thing, while only 11 guys can play on a side at once, most teams rotate in far more than that during the course of the game. Secondly, just because a team returns 15 starters doesn't necessarily mean they're all good. What if the seven who graduated happened to be the seven best players on the team?

2) The presence of a star quarterback and/or running back: We've all heard the old adage "defense wins championships," a theory national champion Florida proved true yet again in 2006, but that doesn't stop us from continuing to be entranced by the more glamorous offensive positions. Without fail, a team that returns veteran star power on offense will be rated higher than a comparable team whose strength lies on defense. Part of that is because it's easier to predict whether a team will field a powerful offense (generally, if they did the year before and they return most of the players, they will again) than it is for defense (where breakdowns and missed tackles can happen to anyone). Teams with a star quarterback or 1,500-yard rusher tend to get more hype than a team with anonymous skill players but a loaded defense -- and voters aren't immune to the hype.

3) A big bowl win: College football is a sport of "what have you done for me lately," and the preseason polls are no exception. Whether a team won or lost its last game the previous season can cause about a six-point swing in the team's rank to start the next season. I have a real problem with this one. Non-BCS bowl games (and even some non-championship BCS games) are often in no way reflective of a team's entire season leading up to it. The teams are coming off a long layoff, some of them are more motivated for the bowl than others and coaches often use the games to experiment with personnel and play-calling in ways they would never consider during the regular season. Therefore, bowl wins and losses -- particularly big upsets -- often create false expectations for the following season.

4) Schedule: In a sport where the schools and conferences make their own schedules, you can't evaluate a team's prospects for the upcoming season without evaluating the favorability of its schedule. How many home games? Does its non-conference schedule include Michigan and Louisiana State or Western Michigan and Louisiana-Lafayette? If it plays in a 12-team conference, which opponents does it miss? The schedule can go a long way toward determining whether a perfectly talented team winds up 9-3 or 6-6. It only makes sense that voters would factor in the schedule when making their preseason prognostications. Doing so, however, raises a thorny question: Is the preseason poll supposed to be a starting-off point or a prediction of how it will look at the end?

The BCS altered its formula in 2006 to place much greater emphasis on the two human polls, thus lessening the possibility of a future split. In instituting the change, however, poll voters now found themselves with more influence than ever before in determining the participants in the national championship game.

Nov. 18, 2006, 11:12 p.m., Ruby Tuesday's, Columbus, Ohio: A couple of my fellow writers and I, having just covered the Ohio State-Michigan "Game of the Century," have gathered for a late-night dinner. Laptop open, I'm putting the finishing touches on my AP ballot, but one issue has me stumped: Who should be No. 2? For weeks, I have resisted in print the notion that, should No. 1 OSU and No. 2 Michigan stage a thriller, the loser should not drop in the polls. The idea of a national-championship rematch seems ridiculous. And yet, with images from the two teams' 42-39 classic still fresh in my mind, I can't bring myself to move another team above the 11-1 Wolverines. Arkansas, the team I had third last week, has won 10 straight games, yet I'm no longer comfortable putting them ahead of 9-1 USC, a team that beat the Razorbacks 50-14 the first week of the season. The Trojans appear to be the next-best candidate but they're only three weeks removed from a bad loss at Oregon State. And Florida, itself 10-1, hasn't looked good in about a month. I poll the table, and everyone is in agreement that, under the circumstances, you've got to keep Michigan No. 2, so I do. It's a decision I'll regret two weeks later when Florida wins the SEC title.

2 of 3