Tip Top 25 in helmets, smaller
                                                    Home

The AP Poll as a National Championship Selector

The AP poll began publishing its weekly top 20 in 1936. It was not the first ranking system or MNC selector that was based on a group's majority opinion, but it was the first such organization that is listed in the NCAA Records Book.

Until the BCS came along, the Associated Press's college football poll was the gold standard for mythical national championships (MNC). Even after the advent of the BCS, people accept an AP poll's #1 team as national champion just as much as a BCS champion, which is why Southern Cal is considered to be co-MNC of 2003 with Louisiana State. The major appeal of the AP poll is the fact that it is a majority opinion. Democracy in action.

The AP poll is generally a good MNC selector because it is the aggregate opinion of many people, but historically speaking, the poll has had a lot of weaknesses as a national championship selector, and some of its choices have been so poor that I do not recognize them as national champions at all.

Pre-Bowl "Champions"

The biggest problem with the AP poll's historical "champions" is that 1936-1964 and 1967-1968, the AP poll ended before the bowl games, and often before every team had even finished the regular season. During those years, #1 teams that lost their bowl games are not national champions at all, at least not in my eyes. Defenders of these "champions" like to say that that's just the way the "system" was back then, but this was not a choice made by the NCAA, but by the Associated Press, which had no official connection to college football. Other organizations counted bowl games all along when naming their champions, and the NCAA itself counted all those bowl games in teams' and coaches' all-time records in their records book. The bowls were never mere "exhibition" games, despite what some like to say.

National championships don't exist at all, of course, except in the eye of the beholder. And if you want to count regular season #1 teams that lost their bowl games as "national champions," be my guest. But such "championships" mean nothing more to me than a post-October "championship." And if you're going to count them in the past, you ought to count them today as well, because nothing has substantively changed except that the AP poll simply chose to do it one way before, and another way now. And that makes their "national championships" very inconsistent from one era to the next.

The following teams finished #1 in the AP poll 1936-1964, then lost their bowl games:

1950: Oklahoma
1951: Tennessee
1953: Maryland
1960: Minnesota
1964: Alabama

These aren't "national champions" of anything other than the regular season. Click on the year to see who ended up #1 in my fixed AP poll for a given season. But #1 teams who lose their bowl games are just the most glaring problem with ending a ranking system before the bowls. If the AP poll had ended after the bowls all along, some of its #1 teams would have been passed up in the final poll despite not taking a loss.

The most famous example is 1947. 9-0 Notre Dame finished #1 in the AP poll, but then 10-0 Michigan stomped on 7-2-1 Southern Cal 49-0 in the Rose Bowl. The AP conducted a post-bowl poll, and sportswriters voted for Michigan, but the AP declared that this poll "didn't count." 11 years prior, in the first AP poll in 1936, 7-1 Minnesota finished #1, but 8-1-1 Pittsburgh would have very likely passed them up in a post-bowl poll thanks to a 21-0 Rose Bowl win over 7-2-1 Washington (whom Minnesota had beaten just 14-7).

These problems are not as egregious as the #1 teams that lost their bowl games, because Notre Dame '47 and Minnesota '36 are legitimate national championship picks regardless, but if you go only by the AP poll's #1 team for your national championships, Michigan '47 and Pittsburgh '36 are left uncrowned. But Michigan and Pitt both claim national championships for those seasons, so obviously they are not sticking with the AP poll as the final say on the matter.

And those cases point to a more basic problem with the AP poll as an MNC selector.

The Highlander Approach: There Can Be Only One

The very nature of the AP poll's methodology makes it a highly limited MNC selector. If the AP poll was a good enough selector, we wouldn't need the coaches' poll to be able to recognize LSU as sharing the national championship in 2003, Nebraska in 1997, Washington in 1991, Georgia Tech in 1990, etc. The point is, sometimes it is not possible to select just one team as THE mythical national champion, nor is it the right thing to do. Yet the AP poll is like the movie The Highlander: there can be only one. Yes, theoretically 2 teams can tie for #1 atop the AP poll, but such a result is extremely unlikely.

The problem here is that the AP poll's real purpose is to rank the teams from 1 to 25 (20 and 10 in earlier polls). Crowning the #1 team as a "national champion" is just a clunky byproduct of that purpose, and not the best way to choose a national champion. The best way to select MNCs would be to have a vote, separate from ranking the teams, in which co-champions were offered up as valid options. For example, the ballot for voters in 1991 might have looked like this:

Washington
Miami-Florida
Washington and Miami-Florida

If 10 writers go with Washington, 15 with Miami, and 55 with Washington and Miami, then the winner is both in a co-championship, without the need for another selection organization such as the coaches' poll. But the way the AP poll is set up, if 79 writers choose 2 teams in a tie for #1, and just 1 writer chooses one of those teams alone at #1, that team is the AP poll's lone national championship selection, and gets the Bear Bryant Trophy. A suboptimal system, to be sure.

Regional Bias

Today, the AP poll apportions its voters fairly by region, but that was not always the case. Through most of the 1950s, any AP writer could vote, and a preponderance of the voters came from the Great Lakes and Eastern regions. Not surprisingly, Southern teams were often disregarded during this time period. 9-1 Ohio State won the vote over 11-1 Georgia in 1942, 8-0-1 Notre Dame and 9-0-1 Army over 11-0 Georgia in 1946, and 9-0 Michigan State over 12-0 Georgia Tech in 1952. Perhaps those choices were fair, but if the vast majority of voters had come from the Southeast and Texas during those years, do you think the outcomes would have been the same?

Regional bias has always been statistically obvious in the voting patterns of football writers, be it voting for teams in the AP poll or for the Heisman Trophy. Most of the voters are writers that cover one particular team and/or conference. And that brings up a related problem...

Limited Viewpoints

Because most of the AP poll's voters cover one team, they are most familiar with that team and its opponents. They may see other games, especially those played on other days but Saturday, but most of their time on gameday is spent with their team, and after that game they have deadlines to meet for their Sunday papers. They don't have time to fairly judge teams outside of their limited purview, even if they are objective enough not to be swayed by bias for their team/region. Limited time is itself another problem...

Limited Time and Lastgamitis

Football writers' busiest workday is Saturday, so they don't have enough time to craft a proper and logical top 25. Because of that, they take rating shortcuts, like just moving teams that lost down from their previous week's top 25. And because they turn in their ballots so soon after the last game is played on Saturday, writers are terribly afflicted by lastgamitis-- judging teams far too much on their last game played. If they had more time, they could step back and get more perspective.

This is much less of a problem in the final poll today, since the bowl season drags on so long, and writers have plenty of time to work on their ratings and reflect on the season as a whole. But in the past, when the AP poll ended after Thanksgiving weekend, there was no time to get any perspective, and lastgamitis ruled supreme.

Shifting Criteria

Another problem with the AP poll is that voters' values and criteria shift like fashion over the years. Math formula ratings, of course, do not have this problem: their criteria is always exactly the same from one era to the next. The only criteria the AP itself has ever given to voters is a reminder to pay attention to head-to-head results, and that has only been in recent years. As such, different voters can have very different criteria. And the zeitgeist-driven values of the majority render one era or season different from another.

The AP poll has been a strong MNC selector ever since it started ending after the bowls, but even in that time it has still voted 2 teams #1 that did not deserve to even share a national championship, and I believe that today's voters would not have made those same choices.

1978

In 1978, the AP poll voted 11-1 Alabama #1 over 12-1 Southern Cal, who beat the Tide 24-14 in Birmingham. Was this because Alabama played a tougher schedule? Far from it. Southern Cal played a vastly tougher schedule, and in fact they played one of the toughest schedules ever faced by an MNC contender. I go into more detail on the issue in my article on fixing the 1978 AP poll, but the point here is that voters today put much more emphasis on head-to-head results than they apparently did then, and so Alabama benefited from different criteria in 1978. The coaches' poll, by the way, got it right in 1978, tabbing USC #1.

1984

The worst choice for #1 in the history of the AP poll was Brigham Young in 1984, the only "national champion" who did not even play a rated opponent. Voters have had plenty of chances to make the same kind of choice (and much better ones) before and since that time, but no other "little big team" has ever been voted #1, or even gotten into the BCS title game, despite a raft of them performing much better than BYU '84 while still playing actual rated opponents. For example, in 1975 Arizona State, then a WAC team, went 12-0 while beating 2 rated opponents, one in the top 10, but they wound up #2 to 11-1 Oklahoma. And we have seen a number of unbeaten WAC and MWC teams in the 21st century who were better than BYU '84, but again, none have even gotten into the BCS title game.

BYU '84 appears to have benefited from a wave of sentiment to finally recognize a minor conference team. Once that was done, writers must have felt that they never had to do it again, and they haven't even come close since.

The AP Poll vs. Other Selectors

Until 1968, when it started counting bowls, the AP poll was worse than human MNC selectors listed in the NCAA Records Book that did count bowl games, such as the National Championship Foundation, the College Football Researchers Association, and the Football Writers Association of America (which gives the Grantland Rice Award to its champions). But since 1968, the AP poll has been much better than the NCF and CFRA, both of whom have made some inexplicably bad MNC selections in that time.

The coaches' poll didn't start counting bowls until 1974, but since that time it has been the only human selector listed in the NCAA Records Book that has been better than the AP poll. The coaches made the same bad choice of BYU in 1984 that the AP poll did, but they correctly voted Southern Cal #1 in 1978, unlike the AP poll. In addition, the coaches made the better choice for #1 in 1991, 1997, and 2003 (details in the linked articles). Those differences aren't too big a deal, because in each of those seasons both teams should be considered co-MNC, but the coaches were correct about which team should be ranked #1. Of course, in 2003, the coaches had no choice: they had to vote for the BCS champion.

The AP poll and the coaches' poll are without doubt the most widely-accepted MNC selectors. Since they started counting bowls, those polls may not have always been correct, but they've been the best, certainly better than any math/computer system.
Home