Welcome to the first ever BizarroMath preseason preview! If you’re greedy for the goods and meh about the math, feel free to skip the specifics and jump right to the juice. You won’t hurt my feelings.
Please enjoy this insightful FAQ.
What is BizarroMath?
BizarroMath (for which I cannot find a good abbreviation)* is my new-and-improved statistical model for projecting the outcomes of college football games played between FBS teams, and the successor to the Super Prediction Action Machine (SPAM).
*Editor’s Note: BM was rejected for obvious reasons, though it does seem fitting for projecting the Iowa offense.
What Was Wrong With SPAM?
SPAM was a byproduct of Iowa’s 2012 and 2013 seasons. You may recall that Iowa went 4-8 in 2012 only to bounce back to 8-4 in 2013. The talking heads found the turnaround remarkable, but I found it unsurprising due to Iowa’s high floor/low ceiling system. SPAM was designed to model Iowa’s close, low-scoring games. I later expanded it to model the rest of the Big 10, and then all of FBS. But, SPAM doesn’t perform as well away from Iowa. These shortcomings were magnified by the impact of COVID on the 2020 season.
The major defects in the system were twofold.
SPAM did not perform direct statistical comparisons between opponents because SPAM could not do opponent adjustments. Instead, SPAM used a ranked weighting algorithm, which tended to pull scores to the middle and shrink spreads. But when you compare teams outside clusters of shared conference opponents, the algorithm is less reliable because the data is not normalized. In other words, SPAM weighted 450 YPG by Ohio State the same as 450 YPG by Arkansas State. This led to badly botched predictions in non-conference games. For example, in 2022, SPAM has Iowa a 0.5 point dog at home vs. Nevada. Nevada is a solid team that could give Iowa fits, but the Hawkeyes should be favored at home.
This could be corrected by applying an interconference adjustment coefficient, but there aren’t enough of data points to calculate one and what little data we have is unreliable. Thus, BizarroMath instead uses averaging and recombination to provide opponent adjustments and normalize all unit data to a national index based on intra-D1 play. This makes heads-up unit comparisons possible.
Also, SPAM calculated win percentage odds based on a regression derived from Vegas spreads. But Vegas sets spread based on getting even money, not fully objective competitive balance. The correct way to model this is to derive a regression based on the algorithm’s accuracy, which I have now done. The new BizarroMath algorithm has been applied to the 2020 and 2021 datasets and I derived a few regression functions. I expected a polynomial regression to be the best fit, but the quadratic coefficient is so small that it drops out of the formula, I don’t even use it. The relationship is almost perfectly linear.
Finally, SPAM’s algorithm was based on the relationship between YPG and PPG, but BizarroMath uses YPP and PPG. The reason for this change is related to my analysis of Iowa’s field position stats in 2021. The bottom line is: if you can’t move the ball, your offense is bad, and even short fields and lots of turnovers won’t change that, so we’re looking at YPP. I am a little wary about this, but we’re going to try it.
BizMarro Math Models The September Games
So let’s get to it, the inaugural pre-season preview for BizarroMath. For each game I provide a brief breakdown of key unit statistics and how BizarroMath sees the matchup, and then I provide my own thoughts. We’ll see who is more accurate, me or the box.
South Dakota State (Sep. 3, 2022)
The Jackrabbits are a salty FCS squad. They went 11-4 last year (5-3 in the Missouri Valley), with wins over FCS No. 2 North Dakota State, FCS No. 14 UC Davis, and on the road at FCS No. 8 Sacramento State and FCS No. 6 Villanova (did you know Villanova has a football team?). They also beat a mutual Iowa opponent, FBS Colorado State, by a score of 42-23 on the road. For comparison, Iowa beat Colorado State 24-14 at home.
BizMa has no prediction here because SDSU is a non-FBS school and BizMa disregards all such data points as statistical anomalies and presumed wins. We are, of course, painfully well aware that we make such assumptions at hazard, but This is the Way.
I think this game winds up being an unnecessarily close nail-biter in which Iowa’s normally solid defense looks a little ragged in an early-season matchup and the offense is tightly wound to avoid turning the ball over and we see little to inspire confidence that anything has improved. But the Hawkeyes salt it away with defense and turnovers, albeit too late to field the second-teamers.
BizzaroMax: Iowa 34, SDSU 20.
Iowa State (Sep. 10, 2022)
The Hawkeyes face what is ideally their first true test when the Cyclones visit Kinnick Stadium. Even with Iowa’s 2.3 point home field advantage, BizarroMath has the Cyclones favored to win this one by 3.51 points, giving Iowa a win percentage of 44.37%. Let’s breakdown what that means practically.
Imagine Iowa and Iowa State play this game 10,000 times. Iowa wins some, Iowa State wins some. Iowa wins sometimes by 50. Sometimes by 1. Likewise with Iowa State. What this data says is that the average Iowa points scored across all those games is 21.23, and the Iowa State points are 24.74, for a spread of 3.51 in favor of Iowa State. This also means that Iowa wins 4,437 of those 10,000 games, or 44.37% of them (by any margin). This of course tells us nothing about which of those 10,00 games we actually get on September 10. It just sets our expectations.
What does the machine see in the Cyclones? Essentially, an overall better version of Iowa - a much better offense offset by a slightly worse defense. The Cyclones’ 2021 offense put up an opponent-adjusted 37.41 PPG, 475.89 YPG, and 7.22 YPP. This is well above the national average. Iowa’s defense can be expected to hold the Cyclones below those numbers, but in a sane universe, the Hawkeyes will need 25-30 points to win this game. On the flip side, Iowa State’s defense is also good, but not as stingy as Iowa’s. This, combined with Iowa’s offense being decidedly NOT good means the aforementioned need to score points is Officially A Problem.
That’s in a sane universe. But we don’t live in a sane universe. We share a universe with El Assico. Anything can happen, man.
I’m pretty sure Matt Campbell has beat every team in the Big 12, yet he hasn’t beaten Iowa, despite having the Cyclones running about as hot as I can remember in my lifetime. How long can that trend hold? I’m going with BizarroMath on this one for two reasons. One, I’m going to this game, which virtually guarantees an Iowa loss. Two, I’m betting against the streak. I know that’s irrational. If you a flip a fair coin 99 times and they’re all heads, the chances of the 100th being heads is still 50/50. But Iowa State has got to break through in one of these years. This is the one. But Iowa State wins in this series are rarely blowouts, so this one is close throughout.
BizarroMath: Iowa State 24.74, Iowa 21.23 (ISU -3.51)
BizarroMax: Iowa State 21, Iowa 20
Side note: As noted, I am attending with an Iowa State friend and we are bringing a third person with us - a beer buddy who went to Old Dominion and has never been to a Big 10 football game. It’s going to be epic. We’re a good time. And we’re looking for tailgating opportunities. Hit me up if you’re game.
Nevada (Sep. 17, 2022)
The Nevada game is where we see the strengths of BizMa emerge. Whereas SPAM has Nevada favored by 0.5 in Kinnick, BizMa’s opponent-adjusted algorithm has Iowa favored by 6.98, with the Hawkeyes projected to score 32.17 to Nevada’s surprisingly high 25.19. This translates to a 61.20% chance of a Hawkeye victory .
The Nevada Wolf Pack went 8-5 (5-3 in conference) and finished third place in the West Division of the Mountain West last year. They run an Air Raid offense that can’t possibly cause any problems. The 2021 Wolf Pack won at Cal to open the season, beat Boise State on the road, and destroyed Colorado State 52-10 on the road. Recall again that then-#5 Iowa squeaked past that same Colorado State team in Kinnick by 10 points.
Iowa should win this game but the projected margin is closer than I expected. Again using opponent-adjusted data, Nevada’s offense puts up a few more points than Iowa State’s (38.04), but doesn’t move the ball as much (435.14), or as well (6.12). Its defense is also less menacing, giving up a lot of points (28.99) and yards (437.21). The math seems about right to me here, but my gut says this is one of those games where it’s 10-7 at halftime, Iowa opens up a 27-10 lead by the middle of the fourth quarter as the opposing defense wears down, and Nevada tacks on a late score to make it closer than it was. I think the machine is too worried about this one and I don’t expect Nevada offense to put up 25 on Iowa’s defense.
BizarroMath: Iowa 32.17, Nevada 25.19 (Iowa -6.98)
BizarroMax: Iowa 34, Nevada 17.
@Rutgers (Sep. 24, 2022)
The Hawkeyes make their first road trip of the season to the east coast to face the Scarlet Knights, who are coming off a quietly respectable campaign. The Knights play a rough schedule in the B1G East and their opponent-adjusted numbers reflect that. The opponent adjustments elevate their raw 18.27 PPG on offense to be worth 21.07, and their 25.64 PPG on defense to be worth 23.88. But statistically, their offense is a little worse than Iowa’s and their defense is nowhere near as good. BizarroMath thus has the Scarlet Knights scoring 13.95 against Iowa and giving up 25.68, for an expected margin of 11.73 in favor of the Hawkeyes. This translates to a 87.19% likelihood of an Iowa win.
It sounds about right, and I haven’t heard anybody prognosticating significant improvement for either Iowa or Rutgers this year. Rutgers played Iowa close the last time the Hawkeyes flew to New Jersey, but that was then and this is now. I expect this one to be uncomfortably close in the first half due to it being Iowa’s first road test on a long flight out east. But Iowa’s defense ought to be able to control this game, and the offense needs to do what it does best: just enough. Rutgers also has All-Multiverse Punting Adam Korsak returning, so I expect a lot of long fields and gross offensive play in this game, resulting in a low-scoring affair.
BizarroMath: Iowa 25.68, Rutgers 13.95 (Iowa -11.73)
BizarroMax: Iowa 20, Rutgers 10
All told, both me and the machine have the Hawkeyes going 3-1 in September with a disappointing home loss to Iowa State. Stay tuned for the October game preview, coming soon!