Writers of Pro Football Prospectus 2008

Most Recent FO Features


» 2018 Free Agency Cost-Benefit Analysis

Is Kirk Cousins the best free-agent quarterback in recent memory? Should Trumaine Johnson or Malcolm Butler have gotten the larger contract? And what makes a free-agent contract good or bad, anyway?

05 Sep 2014

VN: Don't Write Off Week 1

by Bill Connelly

Every college football season follows the same rough calendar, from the one million "FOOTBALL!!!!" tweets that accompany kickoff of the first televised game of the year, to the one thousand "Far Too Early Preseason Top 25" lists that come the morning after the national championship game.

One underrated portion of the college football calendar comes in the six days following the season's first Saturday: Overreaction Season. (Just in case you can't get enough, College Football Overreaction Season is immediately followed by the NFL's National Jump to Conclusions Week.)

This year's Overreaction greatest hits include "Kenny Hill for Heisman," "Todd Gurley for Heisman," "South Carolina is toast," "UCLA is overrated," "Alabama's secondary is still terrible," "Al Golden's done at Miami," "Vandy will be terrible again forever," "USC IS BACK," "NEBRASKA IS BACK," "TENNESSEE IS BACK," "Florida State just doesn't have it this year," and, of course, "Nobody should ever schedule North Dakota State again."

(Granted, the last one is incontrovertible at this point. If you have any interest in winning a football game, make sure NDSU isn't on your schedule, especially if you're Iowa State.)

Some of the above statements will be proven true as the next 13 weeks of the college football season unfold. Most of them, however, won't. And some will be erased from memory by the end of this coming Saturday night.

Overreaction is part of the process, part of the fun of college football season. We have actual, real-life evidence to either support or negate what we were already thinking, and though we're looking at an n=1 situation (or n=14 or so possessions, or n=75 or so plays), we're going to milk this evidence for all it's worth, and then some.

How accurate is this evidence, though? Are Week 1 results as nonindicative to season-long results as we assume? Following a tweet from Tom Gower...

...I decided to look into it a bit. On Wednesday, Brian Fremeau looked at how easy it is (or isn't) to bounce back from a poor first-week performance. By using Adjusted Scores, can we determine whether first-week performances correlate to full-season performance at a higher or lower level than performances in other weeks?

A quick refresher: Adjusted Scores are designed to look at how a given team would have performed in a given week if playing a perfectly average team and receiving a perfectly average number of breaks. They use single-game S&P+ performances and apply them to a general scoring curve for college football. So if the national average for points per game is 28.0 or so (as it has been the last couple of years), then a perfectly average offensive or defensive performance equates to 28.0 points.

As an example, Florida State won last year's BCS Championship game over Auburn, 34-31. Both teams obviously played well enough to beat an average team, however. FSU's Adjusted Score for the game was FSU 28.5, Opponent 19.5. Auburn's was 45.6, Opponent 16.7. FSU's defense performed about nine points better than the average team against the Auburn offense, but they put together a rather average offensive performance (mostly in a lackluster first half). Meanwhile, Auburn's offense achieved far more than most teams had against FSU, and they also did far better than most teams had against the Seminoles' offense.

Because we need a healthy sample of data points, Adjusted Scores are more of an offseason tool, when we can react to what we saw. Used as part of my SB Nation previews, they can show you which offenses and defenses trended up or down as the season progressed, and how much of an impact key injuries had on the team's performance.

And if we compare weekly Adjusted Score performances to a team's full-season averages, we can find out when each team was more "like itself," and when it was better or worse. And if we look at every team, we might be able to get a read on when offenses tend to be better or worse and, using basic correlations, when teams' performances are most and least indicative of full-season performance.

I looked at offensive and defensive performances from the 2012 and 2013 seasons. I could have gone further and done more years, but honestly, I didn't expect to see much in the data, and I very much did not. If I had seen something unexpected (other than the strangely random 2012 Game 14 defensive performances) (for those who had a Game 14), I would have dived into further seasons.

Week /
Correl. to season averages
2012 Offense 2013 Offense 2012 Defense 2013 Defense
Week 1 0.661 0.627 0.577 0.574
Week 2 0.545 0.666 0.626 0.563
Week 3 0.593 0.662 0.634 0.540
Week 4 0.538 0.544 0.556 0.682
Week 5 0.466 0.609 0.483 0.496
Week 6 0.603 0.717 0.448 0.665
Week 7 0.642 0.646 0.629 0.561
Week 8 0.441 0.667 0.553 0.667
Week 9 0.670 0.583 0.616 0.677
Week 10 0.659 0.625 0.538 0.615
Week 11 0.591 0.602 0.523 0.545
Week 12 0.551 0.558 0.575 0.560
Week 13 (Conf Champ Games/Early Bowls) 0.468 0.532 0.421 0.473
Week 14 (Later Bowls) 0.668 0.512 0.068 0.528

I took a few things away from this, most of them with a large grain of salt.

1. Week 1 performances were as indicative of full-season performance as those from any other week, more in three of four cases. Of course, the catch there is that we have to wait a while to find out what the full-season performances are. But if you're worried about South Carolina or super-excited about Texas A&M, you might have reason to be.

2. The 2012 bowl season was strange, at least for defenses.

3. It's not shared in the above data, but offenses performed 1.4 adjusted points worse than their full-season average in 2012 Week 1 and 0.9 points worse in 2013 Week 1. In Week 2, offenses were 1.1 points worse (2012) and 0.7 points worse (2013). Meanwhile, you could see the same phenomenon on the defensive side of the ball -- defenses allowed 0.6 to 1.0 adjusted points fewer in Weeks 1 and 2, then regressed a bit. After that, things got pretty random, but there's at least some evidence there that offenses start to find advantages on defenses as the season progresses.

Week 1 performances aren't particularly random when it comes to the full-season norm. Now we wait to find out what the full-season norm is. Feel free to overreact regardless.

This week at SB Nation

Here are some of my favorite pieces that I wrote at SB Nation this week.

Georgia aced almost every part of its Week 1 test

The Numerical, Week 1: Ole Miss and Wisconsin play 2 games at a time
Bill & Ian overreact to Week 1 at Football Study Hall

Florida State's new questions after the Oklahoma State surprise

The 4 biggest battles in Michigan State vs. Oregon 2014
Week 2 F/+ picks

USC is fun! But when fun plays Stanford, Stanford usually wins
Missouri at Toledo preview

And for fun...

...a Twitter rant.

Felt good.

Posted by: Bill Connelly on 05 Sep 2014

1 comment, Last at 21 Jan 2016, 10:47am by Keshavz1


by Keshavz1 :: Thu, 01/21/2016 - 10:47am

Good to see those tweets and the fight they were doing. Really nice one man. I am looking good areas here.