When we look at how a given algorithm or system of ratings is performing, we tend to look at a single number. My SP+ ratings, for instance, have an average absolute error -- the per-game difference between projection and reality -- of 12.9 points this season in FBS-versus-FBS games, which is decent. Including FBS vs. FCS games, the year's absolute error is 12.3, and it's performing at 55.8% against the Caesars closing line.
Those are all single numbers that tell a single, mostly positive story. But if you home in a bit, you see that its performance, and the sportsbooks' as well, varies pretty significantly by team or conference. And that should obviously affect your betting plans to some degree.
Absolute error for the closing Caesars line involving teams from a given conference (all games, including FBS vs. FCS):
• SEC: 10.2 points per game
• Conference USA: 10.9
• AAC: 11.8
• Independents: 11.8
• Big Ten: 12.3
• FCS teams: 12.9
• Mountain West: 13.2
• MAC: 13.2
• Sun Belt: 13.5
• Pac-12: 14.0
• Big 12: 14.4
• ACC: 14.5
SEC games have been pretty stable overall, with only five matchups varying from the spread by more than 25 points: ETSU's 23-3 win over Vanderbilt (41 points), Kentucky's narrow 28-23 win over Chattanooga (28), Georgia's 56-7 trouncing of UAB (26.5), Georgia's 62-0 destruction of Vanderbilt (26) and Ole Miss' 61-21 win over Tulane (26). Meanwhile, 20 games have finished within five points of the spread.
The ACC also has 14 teams and a similar number of games played thus far. It has also had five games finish with an absolute error of more than 25 points, but only nine have finished within five points of the line. The Big 12, with fewer teams and games, has the same number of big whiffs but only six games finishing within five points.
It is perhaps predictable that SP+ has a similar range of outcomes.
SP+ absolute error and ATS performance by conference (all games):
• SEC: 9.6 points per game, 53.7% against the closing line
• FCS teams: 11.0, 64.6%
• AAC: 11.6, 52.4%
• Big Ten: 11.6, 63.6%
• Independents: 11.9, 46.7%
• Conference USA: 12.0, 48.1%
• MAC: 12.3, 65.6%
• Mountain West: 12.6, 60.4%
• Big 12: 13.2, 56.4%
• Sun Belt: 14.1, 48.7%
• Pac-12: 14.1, 48.9%
• ACC: 14.7, 46.4%
SP+ has been phenomenal in games involving FCS teams, and it's above 50% against the spread for six of 10 conferences, with an absolute error under 13 for six plus independents.
Like Caesars, however, SP+ has whiffed pretty badly on two power conferences in particular: the Pac-12 and ACC.
There's a range within each conference, of course. In the SEC, it's missing Missouri games by just an average of 1.9 points, and it's missing Georgia games by an average of 16.4. But for the combined 26 Pac-12 and ACC teams, its absolute error is under eight points per game for just one (Duke) and 15 points or higher for 12.
For the Pac-12, this may be rectifying itself to a degree. In Week 4, SP+ hit 50% against the spread in Pac-12 games; it came within 8.4 points in five of six games, whiffing only on Oregon State's win over USC. After some super-strange early results -- Washington-Montana and Arizona-Northern Arizona, to name a couple of obvious ones -- it's possible that sanity is beginning to prevail. (At least, that may be the case when USC isn't involved. The Trojans have now dramatically underachieved against projections twice and dramatically overachieved once.)
The ACC is a different story. SP+ virtually nailed the Boston College-Missouri (3.1 absolute error) and Duke-Kansas (3.9) games in Week 4, but it missed five others (Clemson-NC State, Miami-Central Connecticut, Wake Forest-Virginia, Georgia Tech-North Carolina and Pitt-New Hampshire) by an average of 29.6 points.
There seem to be two general causes for these big numbers: inconsistency and/or inaccurate preseason projections.
Five Pac-12 teams and a whopping nine ACC teams have both overachieved and underachieved by at least 12 points in at least one game each, which obviously makes it difficult to figure out where to set the bar. Meanwhile, four teams from these two conferences are underachieving SP+ projections by more than 10 points per game (Washington State -13.4, Clemson -12.3, USC -12.3, FSU -10.7), and five are overachieving by at least 10 (Oregon State +13.8, Georgia Tech +12.1, Syracuse +10.8, Pitt +10.3, NC State +10.2). While the preseason projections hit the mark on plenty of teams, they've whiffed on some in these two leagues. So have the sportsbooks.
With four games in the books for most teams, the weight of preseason projections in the SP+ formula gets diminished pretty significantly moving forward. We'll see if that stabilizes some of these wild swings. Either way, know that if you want to feel a little more at peace with your picks, stick with SEC, AAC, Big Ten and maybe Conference USA games. If you want to ride the lightning, look to the Pac-12 or ACC.
Books and algorithms still playing catch-up on these 18 teams
Four games is a decent sample by college football's standards, and both SP+ and the sportsbooks are theoretically getting a decent read on a lot of teams. But there are still some outliers.
Here are the nine teams that have beaten the Caesars spread at least 75% of the time and are overachieving against the spread by at least 10 points per game:
Rutgers Scarlet Knights: 100% vs. the spread, +17.8 points per game. Greg Schiano's Scarlet Knights are coming off of their first loss and their most impressive performance of the season. They are allowing just 1.1 points per drive (11th in FBS) and 21.3 yards per drive (ninth), and while they couldn't quite reel Michigan in after falling behind 20-3 on Saturday, they came close, shutting out the Wolverines in the second half and falling 20-13. RU ranked 117th in SP+ in 2019, the year before he arrived. Now they're 53rd and still rising.
Bowling Green Falcons: 100% vs. the spread, +17.6 points per game. The Falcons ranked 124th in defensive SP+ in 2019, before Eric Lewis succeeded Brian VanGorder as coordinator. Now they're 67th, having held three consecutive opponents, including Minnesota, to a combined 42 points.
Georgia Bulldogs: 75% vs. the spread, +14.4 points per game. The only time the Bulldogs didn't cover this year was when they allowed South Carolina (+31.5) a garbage-time touchdown to turn a 40-6 cover into a 40-13 loss. The Dawgs have been rampant and just passed Alabama for No. 1 in both SP+ and ESPN's FPI.
Arkansas Razorbacks: 100% vs. the spread, +13.3 points per game. The Hogs were projected to improve, but they've done more than that -- they've already taken down both Texas and Texas A&M, and they're 4-0 for the first time in 18 years. Now they get a shot at fellow overachiever Georgia.
Ole Miss Rebels: 83% vs. the spread, +12.0 points per game. Speaking of big overachievers in big games, Ole Miss heads to Tuscaloosa to test itself against Alabama. The Rebels are listed as 14.5-point underdogs at the moment, but if they continue to overachieve as they have, Bama's got its hands full.
Texas Longhorns: 75% vs. the spread, +11.1 points per game. After a disappointing Week 2 loss to Arkansas, Texas' Steve Sarkisian handed the reins to quarterback Casey Thompson. In two games since, the Horns have scored 128 points and overachieved by 58 points vs. the spread.
Michigan Wolverines: 75% vs. the spread, +10.8 points per game. Jim Harbaugh's Wolverines couldn't quite land the knockout punch against Rutgers and had to take the decision instead; either way, they've passed the tests they've been given to date.
Syracuse Orange: 75% vs. the spread, +10.1 points per game. With a loss only to fellow overachiever Rutgers, Syracuse is 3-1 and has rebounded significantly on defense (43rd in defensive SP+). Not bad for a team that went 1-10 a year ago.
Baylor Bears: 75% vs. the spread, +10.0 points per game. The Big 12 has five of the top 26 defenses in FBS, per SP+, and the Bears lead the way at 11th.
On the flip side, the books are still trying to catch up to how poorly these teams are playing. Here are the nine teams that have beaten the spread 25% of the time at most and are underachieving by at least 10 points per game.
Rice Owls: 0% vs. the spread, -21.5 points per game. The Owls improved to 103rd in SP+ last season but have given away those gains thus far, losing to three FBS teams by a combined 140-24 and giving up 34 points to a low-ranked FCS team (Texas Southern) last week.
Ohio Bobcats: 0% vs. the spread, -21.0 points per game. Ohio was one of a handful of potential MAC favorites heading into the season, but Frank Solich's surprising midsummer retirement has thrown the program for a loop. They lost three games vs. FBS teams by a combined 84 points and fell to FCS' Duquesne.
Vanderbilt Commodores: 25% vs. the spread, -15.6 points per game. Clark Lea's Commodores were projected 106th in SP+ to start the season, easily last in the SEC, but they've fallen to 125th since. They lost by 62 to Georgia only because the Dawgs felt mercy in garbage time.
Ball State Cardinals: 0% vs. the spread, -15.4 points per game. The defending MAC champions needed a full 60 minutes to beat Western Illinois and lost to three FBS opponents by 74 points. Not much of a title encore.
Indiana Hoosiers: 0% vs. the spread, -14.0 points per game. The Hoosiers began the season ranked for the first time in 52 years, but they've lost to two ranked teams by a combined 72-30, and it's possible that each of their next five opponents are ranked as well.
Georgia Southern Eagles: 25% vs. the spread, -13.3 points per game. The Eagles finally beat the spread on Saturday, showing solid fight in a 28-20 loss to Louisiana. It wasn't enough to save Chad Lunsford's job, however.
Clemson Tigers: 0% vs. the spread, -13.0 points per game. The offensive struggles have been well-documented -- the Tigers have scored a total of 39 points against three FBS opponents thus far. And now they're without maybe their best defensive player, lineman Bryan Bresee, for the season.
Florida State Seminoles: 25% vs. the spread, -13.0 points per game. After a rousing near-upset of Notre Dame in the season opener, the bottom has fallen out -- the Noles underachieved against the spread by a combined 56 points and fell to 0-4 for the first time since 1974.
Washington State Cougars: 25% vs. the spread, -12.8 points per game. Nick Rolovich's Cougs beat Portland State in Week 2 to avoid a winless start, but they're 0-2 against the state of Utah and got their doors blown off by a USC team fielding its backup QB. Not great.
At some point, you hit your head on your ceiling or finally find the floor. These 18 teams haven't just yet.