College football projections: Week 14 F/+ picks

Woof.

This was the craziest, silliest week of the college football season. F/+ did terribly … but so did Vegas.

— Bill Connelly (@SBN_BillC) November 23, 2014

F/+ straight-up favs went a miserable 39-20 (66%), but Vegas favs went 40-19 (68%). Sometimes you just applaud CFB for being ridiculous.

— Bill Connelly (@SBN_BillC) November 23, 2014

Through some combination of rivalry effect (if that's actually a thing), crazy weather, and half the teams in the country forgetting how to finish drives, there was a large number of weird games and funky results. Looking back, there were some picks I was a bit queasy about from the start -- Cal covering against Stanford, for instance -- but that's always the case. A lot of last Saturday's results were just strange: Illinois beating Penn State outright, North Carolina whipping Duke, Virginia pasting Miami, Indiana leading Ohio State late in the third quarter, Arizona crushing Utah, Northwestern whipping Purdue -- and I'm semi-comfortable with the numbers missing on those games. Not happy, but comfortable. (And I still think I'm screwing up the home-field adjustment. Those poor 50-59% teams...)

Still, the UNC-Duke result in particular hints at something that's always troubled me with college football ratings: momentum. Or, to use perhaps a more accurate pair of terms: improvement and regression. Thanks mostly to its defense, UNC was dreadful earlier this season, but for the second straight year, the Heels have improved over the home stretch. Duke, meanwhile, had waited until late to pull away from Syracuse, then had lost to Virginia Tech. There was reason to believe regression was underway there.

But the F/+ picks use full-season numbers because it is infinitely safer to do so. Are we really sure Duke is regressing (or losing momentum) because of two games? Are we sure UNC has improved after three super-tight wins (with a blowout loss sandwiched between them)? Full-season numbers drastically favored Duke ... and Duke got stomped.

We're dealing with a sample of no more than 12-13 games in the regular season, so figuring out who really is improving or regressing and who simply had good or bad weeks that were within the acceptable range of whatever their ranking is ... well, that's pretty tough to do, especially in real time.

I say this all in part because there's another game that fits the Duke-UNC profile on Thursday night: TCU-Texas. F/+ sees TCU as a significant favorite because over the course of ~10 games, TCU has been by far the better team. But over the last five games, the Frogs have been all over the map (good, great, solid, great, wholly mediocre), and over the last three games, Texas has looked quite good. The 'Horns are beating bad teams by three touchdowns on the road and whipping solid teams (i.e. WVU) by 17 points at home. Is this a new Texas team? Have the 'Horns simply benefited from being able to outclass three teams from an athleticism standpoint? Was TCU just tired and cold against Kansas? The Frogs have had a bye week to rest and re-establish their previous level of play.

What we think of as momentum is really, really tricky. Our eyes tell us one story (with varying degrees of accuracy), and the numbers tell another. That, plus the general Rivalry Week silliness, makes me pretty much hate this week's picks.

But hey, I'm redesigning how I formulate both my numbers and the F/+ projections in the offseason either way. I shouldn't be afraid of mediocrity, I guess.

Week 14

Here's your weekly disclaimer: the margins are all that matter. The projected scores are for awful, i.e. for entertainment value only.