That we only have 12 opportunities to learn about teams is inconvenient enough. The small-sample college football season requires us to draw broad conclusions about teams when we've only begun to learn about them.
Case in point: NBA teams have already played about 18 games each, and the season started, like, yesterday. In that sport, we can acknowledge we've only begun to figure things out; in college football, the test is almost over. Coaches have been fired because of that sample, and others are about to get a huge raise because of it.
Another harsh truth comes along to make things even messier: teams change.
What you were in mid-September is not what you are in the first week of December.
The Oklahoma that got shoved around in the trenches by a weak Texas is not the same one that showed minimal weaknesses in wins over Baylor, TCU, and Oklahoma State. The confused Alabama offense in a September loss to Ole Miss is far more focused now. The untouchable Michigan defense of the first seven games finished the regular season quite touchable.
The S&P+ ratings I use to supplement much of my analysis use full-season averages. I have yet to find an effective method of giving extra weight to recent games that makes projections more accurate. On average, this works fine. S&P+ is at 54 percent against the spread over the last month, and its win probability figures have exceeded expectations.
Still, some teams change too quickly for the numbers to properly adjust.
Because of Michigan's defensive fade, S&P+ has been overshooting on the Wolverines of late, just as it has undershot on teams like Michigan State and North Carolina.
When the season is over, I post full-season S&P+ numbers along with a Weighted S&P+ figure that signifies, basically, how each team finished the year. It adds context.
Look at 2014, for instance. That No. 6 Ole Miss finished ahead of No. 8 TCU is an eyesore because of the way TCU handled the Rebels in the Chick-fil-A Bowl. But that was the end-of-year version of those two teams; the Ole Miss of the first half of the season was probably the best team in the country. Weighted S&P+ adds context -- Ole Miss' weighted rank was 77th, which tells you how good the Rebels were early and how bad they were late.
So while the official S&P+ projections for this week's games are the more reliable projections, what can a weighted number tell us about these matchups?
Here are the eight teams participating in this weekend's power conference championship games, their S&P+ ratings, and their Weighted S&P+ ratings, which place heavier emphasis on the last eight weeks:
|CFP rank||S&P+||Rank||Weighted S&P+||Rank||Difference|
Michigan State's rating probably stands out to you. The Spartans have looked incredible recently, but very recently -- before beating Ohio State in Columbus and crushing Penn State, Sparty was losing to Nebraska and flirting with the idea of losing to Indiana and Rutgers.
Something else might stand out.
Over the last eight weeks, it might be justifiable to say North Carolina has been better than Clemson.
The Tar Heels' overall ratings have suffered because of results from the first half of the season: their loss to South Carolina, their two games against FCS opponents (including a sleepwalk against Delaware that saw them leading 13-7 at halftime), a less-impressive-than-we-thought win over Georgia Tech, etc.
But while they blew a lead against Virginia Tech and threatened to do the same at NC State, there's no question that they've been awesome of late. They didn't just beat Duke and Miami. They beat the Blue Devils and Hurricanes by a combined 70 points.
At the same time, Clemson's level has dropped. The Tigers were untouchable early, and after surviving Notre Dame, they put up five consecutive performances that were at the 84th percentile or higher. They earned the No. 1 rating in both the committee and S&P+.
Since Florida State, the Tigers have fallen into fourth gear. The win over Wake Forest was fine, but in road games against Syracuse (No. 76) and South Carolina (No. 86), they won by a combined 74-59. Quarterback Deshaun Watson's performance was impressive enough to ensure Clemson was never in danger, but the defense showed clear cracks.
Once Clemson cleared the FSU hurdle, it was natural for the Tigers to let up. But if they are going to get past UNC, the let-up will need to be temporary.
Let's look at the win probabilities for this week's title games.
Again, the full-season, official S&P+ ratings are more likely to be accurate from a projections standpoint. So when you see that Clemson has an 80 percent chance of beating UNC, or that Stanford-USC is a virtual tossup, do not write that off.
But if we were to use the weighted S&P+ ratings to derive win probability, we'd see a different prediction.
|Win probability (S&P+)||Win probability (Weighted S&P+)|
|Big Ten||Michigan State vs. Iowa||Michigan State 62%||Michigan State 50% (-12%)|
|SEC||Alabama vs. Florida||Alabama 76%||Alabama 83% (+7%)|
|ACC||Clemson vs. North Carolina||Clemson 80%||North Carolina 53% (-33%)|
|Pac-12||Stanford vs. USC||Stanford 54%||Stanford 81% (+27%)|
|AAC||Temple at Houston||Temple 51%||Temple 51% (+0%)|
|MWC||Air Force at San Diego State||SDSU 60%||SDSU 80% (+20%)|
|Conference USA||Southern Miss at Western Kentucky||Western Kentucky 78%||Western Kentucky 74% (-4%)|
|MAC||Bowling Green vs. Northern Illinois||Bowling Green 70%||Bowling Green 74% (+4%)|
USC hasn't been that great since Clay Helton took over (or, more fairly, since depth issues began to take their toll in the trenches and at linebacker). The Trojans were strong against Notre Dame and Utah and excellent in the second half of the win over UCLA. But between those performances was a string of mediocrity, beating Cal, Arizona, and Colorado by a combined 17 points and getting run out of Autzen Stadium by Oregon.
Stanford, meanwhile, has continued defensive issues (56th in Def. S&P+) but has sustained a high level on offense (seventh in Off. S&P+).
We have an orderly view of the Playoff, but ...
Oklahoma and the Iowa-Michigan State winner are in, and barring upset, Alabama and Clemson are, too.
But if trends matter, Clemson could have its hands full, and Stanford could win by a large enough margin to make a run at the fourth spot with a UNC win. We might not be done with drama yet.