The 2021 college football season was an odd one for computer ratings.
On one hand, despite all the quirks of 2020's COVID-19 season, the effects of the transfer portal and an early-starting coaching carousel, etc., the game itself wasn't too much harder to project than normal. Using absolute error -- the average difference between projections and reality -- the top 10 performing systems at ThePredictionTracker.com (including only systems that projected every game and not including variations of the line itself) missed the average college football game by 13.0 points per game this season. This average will likely increase at least a hair during the typically volatile bowl season, but considering that average jumped from 12.6 in 2018-19 to 13.4 in 2021, 13.0 represents a nice settling-in toward normality.
Performance against the spread, however, dropped considerably. It has for a while now because the spread itself appears to be getting better and better. While these top 10 systems averaged 52.4% against the spread in 2018, that number sank to 50.9% in 2019, 50.4% in 2020 and a dismal 48.4% in 2021. Among the 33 systems that projected every game, only five ended up over 50% this year; none were over 51%. In 2019, 20 of 44 were at 51.2% or higher.
Fortunes for my SP+ ratings have been no different. SP+'s absolute error was a solid 13.0 points per game as well, and if offered, I'd have absolutely taken that at the beginning of the year because of all the uncertainty. But that was good for only 48.4% against the midweek spread. If you compare its performance to the closing line and include FBS-versus-FCS games (which I don't believe are included at Prediction Tracker), you get up to 52.1% for the season. But that's a far cry from 2019, when it hit 53.4% ATS in FBS-versus-FBS games alone.
The sportsbooks always catch up, and they certainly seem to have absorbed some of the success of the better projection systems out there. But we still have FBS-versus-FCS games and, above all else, we still have win totals. For now.
The books haven't caught up to SP+ on win totals (yet)
Give it time, of course. But flashing back to August's win projections, you could have made a lot of money following SP+'s lead.
If you had bet on every team based on SP+'s win projections, no matter how close the projections and Caesars Sportsbook were to each other, you would have won 60.4% of your bets -- it went 75-48-7 overall. It won 61.9% of its under picks and 59.0% of its overs.
When the disagreement was pretty large, so was SP+'s win percentage.
When it projected a total at least 1.0 win off of the Caesars total, it went 16-5-2 (74%).
When it projected at least 0.6 fewer wins for a team than the Caesars win total, it went 18-3-1 (84%) -- it missed on Notre Dame, Old Dominion and Georgia, pushed on Army and swept the others. When it projected at least 0.6 more wins for a team than the Caesars win total, it went 18-13-2 (58%).
It said to bet the under on every team with a Caesars total of 9.5 wins or higher and went 10-2-1 (81%). It said to bet the over on 17 of 21 teams with a Caesars total of 3.5 or lower and went 10-6-1 (62%).
For a while now, I've had a theory that SP+ and other systems are far better at spotting who's overrated than who's underrated, and SP+ being merely good on extreme overs but spectacular at extreme unders certainly substantiates that a bit.
Best hits
UTEP Miners
Caesars win total (August): 2.5
SP+ projected wins: 4.1
Actual record: 7-5
In our preseason best bets series, UTEP was one of 14 teams for which I had a win total pick.* My theory was pretty simple: The Miners weren't going to be suddenly awesome (they were projected 125th in SP+), but their schedule was so incredibly easy that they should still clear the 2.5-win bar pretty easily.
Sure enough, Dana Dimel's squad went 6-0 against teams that finished 109th or worse in SP+, 1-5 against everyone else and cleared the bar with ease. Granted, they were also better than projected -- 90th in SP+, thanks primarily to a sound and aggressive defense -- but they primarily beat the beatable teams, and there were a lot of them on the schedule.
*The 14 picks went 8-5-1 overall, 60.7%, which means that the 116 other picks I didn't make best bets had almost exactly the same win percentage. I'm great at this.
BYU Cougars
Caesars win total: 6.5
SP+ projected wins: 8.1
Actual record: 10-2
This one felt like an overadjustment on the books' part. BYU lost almost as much of 2020's production as anyone in college football and faced a schedule that featured seven power conference opponents. But the Cougars were still projected 31st in SP+, and only three of those P5 opponents ended up as top-40 teams. They started quickly with a series of close wins -- they beat Arizona, Utah, Arizona State and USF by an average of 28-19 -- and rolled to 10-2 from there despite ranking only 42nd overall.
Northwestern Wildcats
Caesars win total: 6.5
SP+ projected wins: 4.9
Actual record: 3-9
Average returning production in college football in 2021 was off the charts, but that didn't apply to Pat Fitzgerald's Wildcats. The defending Big Ten West champs were last in returning production by a significant margin, and the inexperience showed. SP+ projected them to fall to 76th, and that turned out to be an overestimate -- they ended up 107th (69th on defense, 118th on offense and 129th on special teams).
LSU Tigers
Caesars win total: 8.5
SP+ projected wins: 7.1
Actual record: 6-6
For any preseason radio hit I did, I called LSU one of 2021's greatest wild cards. Teams capable of going 15-0 in one season (as the Tigers did in 2019) aren't supposed to be capable of plummeting to 5-5 the next (as they did in 2020). The Tigers headed into this fall with massive upside and scary downside, but SP+ took what we saw in 2020 pretty seriously and projected them in the mid-20s. As it turns out, that was giving them too much credit. They struggled mightily with injuries and fell again to 61st.
Ohio State Buckeyes
Caesars win total: 11
SP+ projected wins: 9.7
Actual record: 10-2
The 2019 season, which saw Ohio State, Clemson and LSU all reach 13-0 before losing amongst each other, was the general exception, not the rule. Not all seasons feature quite as much oddity as 2021 -- which ended up giving us only two 12-0 teams (Georgia and Cincinnati), one of which made it to 13-0 -- but it's a general rule of thumb that even most of the best teams lose. Ohio State was no exception. After winning 25 straight non-CFP games going back to 2019, the Buckeyes were a more normal great team, dominating plenty but also losing twice. They were one of four teams with a Caesars win total of 11 or 11.5. All four hit the under.
Worst misses
On the flip side, here are the five teams on which SP+ and Caesars disagreed the most and SP+ whiffed.
Florida International Panthers
Caesars win total: 4.5
SP+ projected wins: 6.1
Actual record: 1-11
Wow, did things fall apart for Butch Davis. After going a combined 15-11 in 2018-19, his Panthers slipped to 0-5 last season while taking a particularly hard hit from COVID. SP+, still taking those previous seasons into account, projected improvement toward at least merely bad play, ranking them 110th in the preseason. They ended up 125th. They beat FCS newcomer Long Island in the season opener, then lost 11 games to FBS teams by an average of 42-18.
The offense hit some deep shots here and there, the defense couldn't stop anyone all year, and Davis said the school is "sabotaging the program" on the way out the door.
Tulane Green Wave
Caesars win total: 5
SP+ projected wins: 6.1
Actual record: 2-10
It wasn't a total surprise that Davis couldn't get things going again at FIU, but the level to which Tulane fell off course this year was jarring. The Green Wave reached bowls for three straight years under Willie Fritz and began 2021 by nearly upsetting Oklahoma. But they beat only Morgan State and USF and lost to all 10 top-80 teams they played. Now, the schedule played a role in their demise -- it featured trips to Oklahoma and Ole Miss, a visit from Cincinnati and two teams that finished worse than 76th -- and they fell only to 89th overall, which isn't FIU-level horrible. But they were projected 55th. That's still a big miss.
Notre Dame Fighting Irish
Caesars win total: 9
SP+ projected wins: 7.9
Actual record: 11-1
SP+ was nearly perfect with its big "under" picks, but it missed on the Fighting Irish. It was low on Brian Kelly's (now former) team, both because of their low returning production figures and because their rating fell quite a bit late in 2020. It projected them 18th, but they are currently seventh, and a combination of that overachievement and underachievement from many of their opponents -- Virginia Tech, USC and North Carolina, in particular -- gave them an opportunity for another huge season. They survived some early and unexpected close calls but looked spectacular down the stretch.
Colorado State Rams
Caesars win total: 4.5
SP+ projected wins: 5.6
Actual record: 3-9
Steve Addazio's Rams had one of the strangest seasons in FBS. They began the year with losses to FCS heavyweight South Dakota State (forgivable, honestly) and Vanderbilt (less so) but surged, stomping Toledo, San Jose State and New Mexico and giving Iowa fits. They were positioned to overachieve expectations, but they suffered a terrible special teams miscue at the end of a loss to Utah State (after which Addazio seemed to primarily blame the players and not the coaches), and things completely fell apart from there. They ended the year on a six-game losing streak, and Addazio was fired.
Florida Atlantic Owls
Caesars win total: 7
SP+ projected wins: 8.0
Actual record: 5-7
I felt pretty good about Willie Taggart's Owls, enough to make "FAU +1200 to win Conference USA" another of my preseason best bets. SP+ projected them 75th, and that's where they still stood after a 3-2 start. They made it to 5-3 but lost each of their last four games by an average of 34-16. The defense continued to play pretty well, but the offense was beset by turnovers and inconsistency. Taggart is now just 10-11 in two seasons since replacing Lane Kiffin.