One of the longstanding traditions of the college football poll system is ranking teams in order of losses. There obviously is a divide between power conferences and non-power conferences, as any Boise State fan can tell you, but pollsters are generally reticent to put Team A ahead of Team B if Team A has more losses than Team B does.
To illustrate how this works, I looked at AP poll finishes as recorded by the excellent College Poll Archive. I looked at seasons with the 12-game scheduling format we have now, which are 2003 and 2006-14.
A quick history note: those old enough may remember that the NCAA did a test run of the 12-game schedule in 2002-03 before making it permanent in 2006, but teams could still pick up an extra regular season game via "preseason" "kickoff" games in 2002. For instance Florida State and Iowa State played in the 2002 Eddie Robinson Classic, which allowed them each to play 13 games prior to their bowls without having played a conference championship game. The NCAA did away with those games after 2002, making 2003 the first year with a scheduling system equivalent to today's.
Here is a graph of how many losses the teams in each AP ranking slot had on average in the ten seasons I looked at:
Things start to get a bit messy in the high teens thanks in large part to the effect of non-power conference teams. Here's the same graph again, but this time with only power conference teams included. The form of the trend is a bit neater when you don't have two-loss MAC teams ranked in the 20s and so forth.
I included all BCS auto-bid leagues (and the 2014 Power 5) plus Notre Dame as power conferences except the AAC in 2013. I left out the AAC of that year because, while it had an automatic BCS bid for its champion, the AP voters treated it like a non-power conference in the way they placed its teams. The voters didn't really treat the Big East as a true equal to the other BCS auto-bid leagues after the ACC's 2004-05 raid, but it wasn't nearly as bad as the AAC's treatment in '13.
It still goes fairly linearly, but it levels off in the 20s because nearly all of the power conference teams ranked in the 20s had four losses. There are two spikes at 22 and 24, which are the only spots where five-loss teams have ended up ranked, and it dips down at 25 thanks in large part to three-loss Cincinnati having ended up there in 2011. The spike in the 2 slot is mostly due to the custom of pollsters being reluctant to penalize one-loss national championship game participants for picking up a second loss in that game.
Speaking in terms of power conference teams, finishing undefeated is basically a guaranteed national championship—unless you're on probation (sorry, 2012 Ohio State) or we somehow end up with all of the Power 5 conference champions being undefeated (which is incredibly unlikely).
For one-loss teams, you're probably going to end up in the top five unless you started the year unranked and unexpectedly finished with only one loss thanks in part to a weak schedule (2006 Wisconsin, 2007 Kansas) or you're in the old Big East (2006 Louisville, 2009 Cincinnati). Having two losses is usually good for a top ten finish unless you did something like follow up an iffy schedule with a bowl loss (e.g. 2008 Texas Tech, 2010 Michigan State, 2013 Ohio State and Baylor) or, again, you were in the old Big East (2006 Rutgers, 2012 Louisville).
Things begin to get messy with three and four losses, but there is an inflection point at the 17/18 spot in the poll:
With three losses, you'll be anywhere from 6 to 20, probably in the 11-17 range, as long as you're not 2011 Cincy. Four losses can be good enough for a top 15 finish, but high teens to 20s is far more likely.
Only five teams with five losses have found the final poll in the ten years I specified. Two finished at No. 22: 2010 South Carolina and 2014 Auburn. Three more found the poll at No. 24: 2003 Florida, 2009 Clemson, and 2012 Michigan. The Palmetto State teams were otherwise four-loss teams that lost conference championship games. The '03 Gators and '12 Wolverines only lost to teams ranked in the final poll, and the former beat eventual national champ LSU*. Last year's Tigers had four losses to teams ranked in the final poll while playing in toughest SEC West ever, and one of the losses was in overtime.
Keep these parameters in mind as we head into the thick of season preview time and you're looking at schedules to project how many losses you think any given team will have. It will come especially into play with this year's impending SEC West logjam, where a really good team is going to have four losses for no reason other than playing in that division. Maybe that team is really good, but chances are decent that it'll end up ranked outside the top 15.
*Yes, Ron Zook beat a national championship Nick Saban team IN SABAN'S HOME STADIUM and then lost to a team that lost to Maine the very next year, which is why he's the most randomest of head coaching random number generators.