Home / Columns / TWER presents Mac Mirabile’s Predict-O-Nomics

TWER presents Mac Mirabile’s Predict-O-Nomics

Alice can't believe it - we've got another model*... and you might actually have a shot at with this one.

For the last seven years, I’ve put my Auburn economics degree to work by developing gambling models.  These picks and the subsequent discussions are based on mathematical models developed explicitly for The War Eagle Reader. Using historical data since 2004, I’ve come up with a couple of models (used below) to predict the margin of victory by the home team (similar to the Vegas line), and the total points scored in the game (similar to the Vegas totals). The predictions from these two models are combined to predict the score.

These predicted outcomes may correspond very closely to Vegas’ prediction for the game with one key difference: in Las Vegas, the odds are set to balance an equal amount of money on each side of the line.  This ensures that the “house” always gets paid. However, it also tends to exploit irrationality in the gambling market.  For example, the opening line may be set too high in order to capitalize on the irrational exuberance of Alabama fans who always expect their team to cover the spread regardless of the situation.  The more rational bettors may make offsetting bets against Alabama covering; this way the house optimizes its take by understanding how bettors tend to think about certain teams.

That said, the Vegas line is a pretty good predictor of a game’s actual outcome.  In a large enough sample of games, it is, in fact, unbiased, which means it’s difficult or impossible to get rich by exploiting any systematic failures or asymmetries in the line – they simply don’t exist.

The following predictions are built on dozens of carefully selected measures used to describe both the margin of victory and the total points we’re likely to observe in a given game.  If one was so inclined, he could use these predictions to place bets.  In fact, through the first three weeks of games, this model is 77-57 against the spread in all lined games and 14-10 against the spread in games involving SEC teams.

I will get into more detail in the coming weeks, but the model contains comparisons of the season-to-date offensive vs. defensive numbers of both teams, as well as incorporating conference game controls, rivalry game controls, AP Poll ranking, stadium size, and win/loss streaks.

And now… the predictions:

UAB @ Tennessee (-14, 51) — 11:20AM CT
The Tennessee Volunteers look to break their two-game losing streak at home when they face an unranked UAB team. The model suggests these two factors cancel themselves out — basically an unranked Volunteer squad with a two-game losing streak would play a fairly even game against an unranked opponent. I buy that. The Vols do receive a 3-point margin based on how their 7-6 record compares to UAB’s 5-7 record from last season. The Vols lose a whopping 9 points due to their last-in-the-SEC-returns team play, which will be compounded by UAB’s No. 4 nationally ranked net punting (44.5 net). The only reason the Vols will win this game is because of the 100,000+ fans packing Neyland Stadium for a rare view of a Tennessee win this season. Prediction: Tennessee wins by a touchdown over the Blazers, 34-27.

Alabama @ Arkansas (+7.5, 54.5) — 2:30PM CT
The Vegas line on this game has Alabama winning by at least a touchdown, which seems optimistic to me if you’re a Razorbacks fan. I’ve got Alabama by 11, though the Auburn man in me would love to be wrong. Alabama’s 17-game winning streak is worth 2 points by itself. Arkansas’ relatively small stadium (by SEC standards) also costs them about 3 points. Anyone who went to last week’s game at Jordan-Hare Stadium knows what 87,000 fans praying for a miracle can accomplish. The special teams game play also heavily favors Alabama. Prediction: Alabama wins, 31-20.

Kentucky @ Florida (-13.5, 51.5) — 6:00PM CT
Defending SEC East champion Florida hosts a high-scoring Kentucky team ranked No. 11 nationally at 44.3 PPG. I’m going to preface this discussion by noting of all the picks this week, this is the one I believe in the least. The model suggests that Florida is ranked for a reason and Kentucky is not. Florida has enjoyed one-loss seasons three of the last four years, while Kentucky is a perennial 8-5 or 7-5 team over the same period. Teams tend to perform consistent with the level of talent in the program, and comparing recruiting classes over the last five years, this should be a very one-sided game. Despite how poor Florida’s offense has looked to date (No. 92 nationally with only 319 yards of total offense per game), they are managing to put up some points (No. 32 nationally with 34.3 PPG). Both Florida and Kentucky have very efficient defenses so far, the biggest difference here is strength of schedule: Florida’s opponents are 4-1 against other teams, while Kentucky’s opponents are 1-5 against other teams with that lone win against lowly directional school Eastern Kentucky. Look for the Gators to slow down the Wildcats while picking things up on offense themselves. Prediction: Florida 39 – Kentucky 18.

Georgia @ Mississippi State (-1, 47) — 6:00PM CT
Both sets of Bulldogs enter this game having lost games they could have won (Georgia vs. Arkansas, Mississippi State vs. Auburn), and both look to avoid going 0-3 in the SEC. Such starts are old hat for Mississippi State, with the old reprieves occurring when Vandy is on their schedule. Georgia, however, is going to be ready for this game and ready to capitalize on the unfortunate yet inevitable turnovers by Mississippi State. Neither team has a clear advantage in any of the statistical matchup metrics according to the model in what should be a very close game. Look for MississipPi State to rebound from last week’s five turnover game and keep this one close. Prediction: Georgia 26 – Mississippi State 23.

Fresno State @ Ole Miss (-2.5, 54) — 6:30PM CT
Just when you thought things couldn’t get any worse for Ole Miss, Fresno State is coming to town. Losses to Jacksonville State and Vanderbilt have shattered the Rebels’ confidence, and the Bulldogs from the WAC are going to make things even worse. Neither of these teams has much going for it offensively or defensively, though the stats matchup favors Ole Miss slightly. That said, Fresno State has a special teams advantage and will be looking forward to capitalizing on Rebel turnovers. Prediction: Fresno State 30 – Ole Miss 27.

South Carolina @ Auburn (-2.5, 46) — 6:45PM CT
The Vegas line on this game looks dead-on. I’ve got Auburn by a field goal, and this should be a tight contest all game long. Both teams come in ranked similarly, both sporting mild winning streaks. Auburn’s AP ranking of No. 17 is worth 3 points, but South Carolina’s ranking of No. 12 costs us 5 points. The fact that South Carolina is allowing 267 passing yards per game means Auburn’s current 201-per-game average has room to expand. This is worth one touchdown according to my model. However, Auburn’s 260 yards per game rushing will face a stiff challenge against a Gamecock defense allowing on 60 yards per game: This is going to cost Auburn a couple of scores. That said, Auburn’s offense is going to be better than South Carolina’s defense and will find a way to win. Prediction: Auburn 23 – South Carolina 20.

West Virginia @ LSU (-8, 42) — 8:00PM CT
Death Valley at night. The LSU Tigers are difficult enough to beat during the day, and West Virginia is going to need to play its best game of the season to keep this one close. The Tigers’ No. 32 rushing offense is facing the Mountaineers’ No. 8 rushing defense, which is a push with no sizable advantage for either team. West Virginia’s passing game and LSU’s lack of one moves the predicted margin of victory in favor of the Mountaineers by 1 point. The Tigers make up for this in special teams where their No. 5-ranked punt return team and No. 13-ranked kick return team will keep the Tigers consistently playing on a short field. Prediction: LSU 30 – West Virginia 17.

Next week, I’ll be refining the models to take into account strength of schedule, which prior to week 4 is difficult to assess. Time permitting, I’ll also create some new predictive variables taking into account performance relative to time of possession.

Mac Mirabile is a 2002 graduate of Auburn’s economics and journalism departments. During his time at Auburn, he was a copy and photo editor with The Auburn Plainsman. He has a master’s in economics from UNC-Chapel Hill and has written numerous academic publications on college football, the NFL and gambling markets. He can be reached at [email protected]

* Totally joking, Al… we’ll never let you go, graduation be damned… besides, we couldn’t think of anything else to use here besides a pie chart or something.

About Mac Mirabile

Check Also

les

The AU Wishbone Podcast: Interesting, Coach!

Van and John return with their THIRD episode of the week (!!), closing the books …