Tip Top 25 in helmets, smaller
                                                    Home

How to Rate College Football Teams

Head-to-Head Results

It may seem silly that the AP poll reminds its voters of something so obvious as to pay attention to head-to-head results, but looking at the polls, they obviously need to do more than remind them. Perhaps they should italicize the reminder, put it in boldface, and highlight it in red. Maybe send helpful instructional videos explaining the concept. Or they could simply send voters a link to this helpful guide!

The simple rule is this: if two teams have a comparable record, and one has beaten the other, the winner should be ranked higher most of the time (let's say about 90% of the time, at least). And by "comparable record," I don't mean straight record, but relevant record.

Michigan vs. Wisconsin in 2006

Let's go back to that 2006 example from the end of the previous section. Michigan defeated Wisconsin that year 27-13-- a decisive outcome--but Wisconsin is rated #7 in the final poll, and Michigan is #8. That is because Wisconsin has a better straight record (They were 12-1 and Michigan was 11-2). But do they have a better relevant record? Well, you know I wouldn't have brought them up if they did.

The fact is that Michigan's only 2 losses came to the #2 and #4 teams, both ranked higher than either Wisconsin or Michigan. So when comparing Michigan to Wisconsin, those 2 losses are irrelevant, and you can think of Michigan as 11-0 and Wisconsin as 12-1, having lost to Michigan. Or you can simply look at the fact that Michigan lost to #2 and #4, and defeated #7, #17, and #24, and ask yourself where they should logically be ranked. Obviously they should probably be ranked #5 or #6.

The point is this: Wisconsin has a better straight record, but Michigan has a better relevant record. Like Arkansas in the same season, AP poll voters basically punished Michigan for playing a much tougher schedule than Wisconsin did. Wisconsin was rewarded for not having to play Ohio State that year, and for getting Arkansas as a bowl opponent, whereas Michigan got USC-- who beat Arkansas that year 50-14! Michigan also defeated 3 top 25 teams that year, Wisconsin only 2. And again: Michigan's win over Wisconsin was by 14-- decisive.

This is such a no-brainer that I would go so far as to say that any voter who had Wisconsin over Michigan (apparently the majority) should not even be voting. It is that clear a case. And it casts every other choice the AP poll makes as dubious at best. A farce at worst.

Mitigating Factors

Of course, sometimes head-to-head results are not so clear. For the sake of argument, let's say that Michigan's losses in 2006 came to the #2 and #15 teams. That would give them a loss to a lower-ranked team, but their win against Wisconsin balances it out. The two teams would essentially have the same relevant record, with the head-to-head result breaking the tie in favor of Michigan. But now let's say that that win over Wisconsin was Michigan's only win over a rated team, while Wisconsin had 3 wins over rated opponents. And let's say that Michigan only beat Wisconsin by 3 at home. If all of that happened, then you could logically rank Wisconsin over Michigan, and probably should.

But that is all hypothetical. For a real world example, consider the famous case of Florida State and Notre Dame in 1993. In a minor controversy, 12-1 FSU was voted #1 over the team that beat them, 11-1 Notre Dame. Why? Three reasons. First of all, FSU played a tougher schedule than did Notre Dame. Secondly, FSU performed better than did Notre Dame (in other words, FSU stomped on almost all of their opponents, even rated ones, more than Notre Dame stomped on theirs). Lastly, Notre Dame's win over FSU came at home, by 7, and FSU drove into the Notre Dame red zone and got one shot at the tying touchdown in the closing seconds. It was not a decisive win. There are actually other reasons FSU was voted #1, but those are the three valid, logical reasons why you could rate FSU #1 for 1993.

But note that I am not saying that FSU should be rated #1 for 1993. I am just saying that logically, they can be. Either FSU or Notre Dame is a valid option for #1. So is 11-0 Auburn, by the way. But they were on probation, did not play a bowl game, and finished only #4 in the AP poll. I myself would have voted for Notre Dame, but it is a very tough case, and there is no right or clearly logical answer.

The Drumbeat for Head-to-Head Results

In recent years, there has been a growing drumbeat for head-to-head results to be taken into better account by AP poll voters. Leading the charge are Stewart Mandel and Andy Staples of Sports Illustrated (SI's last two voters as of the 2009 season) and Hugh Falk of pollspeak.com, a website I recommend. Falk has made it his mission to publically flog bad individual voters and laud good ones. He combs through hundreds of ballots to find the voters whose feet need to be held to the fire, God bless him. I am a fan of the efforts of all three, but I think that even they could use a little sharpening when it comes to head-to-head results.

Stewart Mandel's Big Catch

Stewart Mandel is my favorite college football writer, and the primary reason I ended up shifting from being a Sporting News man to a Sports Illustrated man when it comes to college football. In 2007, when writing about his final AP poll ballot, Mandel focused heavily on the head-to-head results of these top ten teams: 11-2 West Virginia, who beat 11-3 Oklahoma, who beat 12-2 Missouri (twice!), who beat 12-1 Kansas. And that is the order in which he ranked them. 

But just looking at the straight records, you know that the AP poll did not rank those teams in that order. And indeed, the AP poll ranked them like this: #4 Missouri, #6 West Virginia, #7 Kansas, and #8 Oklahoma. Congratulations are due to Mandel for being among the very few that year to look at these teams closely and pay attention to head-to-head results. Oklahoma was the Big 12 champion, defeated Missouri twice, and dominated them both times (winning by 21 in the Big 12 title game). How many times did poll voters need to see the two teams play to understand that Oklahoma was just a better team?

The Ones That Got Away

Unfortunately, Mandel apparently did not spend nearly as much time and effort on the bottom part of his ballot as he did on the top. I suspect, from looking at the AP poll over the years, that this is true of most, if not all, voters-- that they put more effort into the top 10 than they do #11-25. In any case, that same season, Auburn defeated Florida, who defeated Tennessee, and all three teams had 4 losses (and comparable relevant records too). But Mandel, along with the AP poll at large, ranked those teams in the exact opposite order (AP had Tennessee #12, Florida #13, and Auburn #15). That is actually a much clearer case for head-to-head than the 4 teams in the top ten, since they all had 4 losses.

But that's not all. In the same season, Oregon State defeated Oregon, who defeated Michigan, and again, all three teams had 4 losses. And yet again, Mandel and the AP poll ranked these 3 teams in the exact opposite order (AP had Michigan #18, Oregon #23, and Oregon State #25).

Why play the games at all if the results are just going to be disregarded so very often? Along with taking strength of schedule properly into account (i.e. looking at the relevant record rather than the straight record), recognizing head-to-head results is where the AP poll is in greatest need of improvement. And that includes even its best voters.

Conclusion

I'll end this section with a sharpening of Hugh Falk's head-to-head criteria. Falk is doing a good thing at pollspeak.com, but his approach to head-to-head results is a bit too simplistic. His mantra is that if two teams have the same record, and one defeated the other, the winner should be ranked higher. The problem is this: just because two teams have the same straight record does not mean that they have the same real (or relevant) record.

For example, after November 21, 2009, when 8-3 Mississippi defeated 8-3 LSU at home 25-23, Falk got on voters' cases for voting LSU higher than Mississippi. But the problem is, voters should have been ranking LSU higher, because despite the fact that both were 8-3, LSU actually had the better record. LSU's other 2 losses were to Florida and Alabama, then the top two teams in the country. Mississippi also lost to Alabama, but their other two losses came to two lower rated teams (South Carolina and Auburn). The only reason they were both 8-3 is because LSU had played Florida, and Mississippi had not. 

Because of Mississippi's two upset losses, they were in effect two games behind LSU before the two teams even played. After Ole Miss beat LSU, they were still effectively one game back. And let me reiterate-- Mississippi's win was by 2 points at home. Not at all decisive. Certainly nowhere near enough to dismiss LSU's better overall record. Of course, this all became moot when Ole Miss lost at Mississippi State the next week.

Yes, the AP poll definitely needs to get better-- much better-- at recognizing head-to-head results. But in the messy world of wildly varying college football schedules, things are rarely what they seem on the surface. No simplistic rule can substitute for carefully looking at teams' records-- their real and relevant records-- before bowing to head-to-head results.

Next: Strength of Schedule

Sections that follow "Strength of Schedule":
Other Criteria: Performance, Improvement, and Common Opponents
How to Rank the Top 25 Teams
How Not to Rank Teams

Home