I'm taking a brief sabbatical from the quarterback question (I'll have the splits for BCS versus non-BCS teams shortly) and focusing on an old reliable: close games. What I want to know is: How often (at least in 2006) did the 'better' team win a close ball game (decided by 8 points or less)? Now defining better will get you in a ton of trouble, but I've decided to try and do it anyway. For the purpose of this analysis, I'm looking at each team's record outside the game in question. Now of course, this ignores schedule strength and a host of other variables including when the game was played. Some teams improved dramatically over the course of the season (Arkansas and Wake Forest) and some teams fell apart (Clemson). Still, its a quick and easy measure and it can tell us a little something about team strengths. If the teams have the same record outside of their close game then it gets ignored. For example, in a Week 1 game that broke the TV ratings record set by M*A*SH, Buffalo beat Temple 9-3 in overtime. Buffalo finished the season 2-10, while Temple finished 1-11. Outside of that game, both teams were 1-10, so they are treated as being roughly equal and the game is not counted. In addition, the team's record outside of the close game must be more than 1/2 a game different, in order to adjust for the varying season length of each team's schedule which can include bowl, conference title, and/or Hawaii games. For example, in mid-October, Alabama beat Ole Miss 26-23. Alabama finished the season 6-7 after losing their bowl game. Ole Miss did not play in a bowl game and finished 4-8. Outside of their game against each other, Alabama was 5-7 and Ole Miss was 4-7. Since the difference is only 1/2 game, this game is also ignored and the teams are treated as equals.
In 224 games decided by 8 points or less. where there was at least a game difference between the teams outside of the close game, the 'better' team had a cumulative record of 139-85. The 'better' team won a close game a little more than 62% of the time in 2006. That seems like a pretty strong majority, and it would be in a presidential elections. But look at it the other way. The weaker team has a 38% chance (better than 1 in 3) of pulling off the upset. Most underdogs would gladly take those odds.
Finally, here's a theoretical example. Say your a pretty good team in a BCS conference. Let's go ahead and assume your the best team in your BCS conference. But your conference is very strong (say like the SEC in 2006) and you have to play three teams, that while a notch below you are still pretty good. And each of those games turns out to be a nail-biter. The odds that you win everyone of those games is just 23.9% (.62 ^3). And what about that road game against a decent team where you took them lightly and they played you tight deep into the fourth quarter. Odds that you win all 4? 14.8% (.62 ^4). The point I'm trying to hammer home yet again, is that close games have a lot to do with lady luck. Even when you are the better team, the fates can intervene and hand you a close loss.