By Ross Benes

Leading up to bowl season each year, there's so much turnaround among college football coaches that it feels like there's a hiring or firing every other day. And whenever a bowl-eligible team announces a coaching change at the end of the regular season, the narrative behind their bowl contests shifts as commentators start speculating how teams will "respond" to their interim leader. But when it comes to accurately predicting bowl game results, is there any noticeable difference between the performance of teams led by their regular head coach and those led by a temporary fill in?

To find out, we compared bowl games' expected results against their actual outcomes. We used Ed Feng's The Power Rank to measure anticipated scores, which compute each team's expected margin of victory against an average opponent and factors in strength of schedule, margin of victory and offensive and defensive statistics adjusted by opponent strength. We then compared the difference between teams' scores on The Power Rank to the outcome of their bowl game.

For example, leading up to last year's Outback Bowl, Wisconsin had a score of 22.97 on The Power Rank and Auburn had a score of 30.66, which means Auburn was favored by 7.69 points heading into the game. But Auburn ended up losing by three points, so the game missed its expected outcome by 10.68 points. We carried out this exercise for every bowl game this past decade, and split the results into interim and non-interim coached games.

To create a list of interim-coached games we relied on collegefootballpoll.com's coaching-change tracker. The sortable table below shows every bowl game from 2005-14 that featured one interim coach. "Year" corresponds to what regular season led to the bowl. (To limit confusion games pre- and post-New Year's are included under the same "Year" if they followed the same regular season.) "Expect" is how much the winning team was expected to win/lose by according to their score on The Power Rank. "Result" is how much the winning team actually won by. The next column is the difference between these two scores. "Predict" tells whether or not the winner was correctly predicted. "Rk" ranks each game by how well it conformed to expectations. (For example, the Rose Bowl following the 2012 season in which Barry Alvarez interim coached Wisconsin ranks first because the game unfolded as predicted with Stanford winning by six points. The GMAC Bowl following the 2008 season ranks last because a team expected to lose by about a field goal, Tulsa, ended up winning by 32 points.)

Year Winner Loser Expect Result Difference Predict Rk
2012 Stanford Wisconsin 6.58 6 -0.58 y 1
2011 Texas A&M Northwestern 11.72 11 -0.72 y 2
2007 Texas Christian Houston 7.76 7 -0.76 y 3
2005 Clemson Colorado 8.08 9 0.92 y 4
2007 Penn State Texas A&M 6.03 7 0.97 y 5
2012 Cincinnati Duke 15.18 14 -1.18 y 6
2014 Southern California Nebraska 4.25 3 -1.25 y 7
2007 Brigham Young UCLA 2.41 1 -1.41 y 8
2012 San Jose State Bowling Green State 11.06 9 -2.06 y 9
2012 Arkansas State Kent State 1.00 4 3.00 y 10
2009 Texas Tech Michigan State 13.15 10 -3.15 y 11
2006 Oklahoma State Alabama -1.86 3 4.86 n 12
2006 Central Michigan Middle Tennessee State 11.83 17 5.17 y 13
2009 Central Michigan Troy 8.28 3 -5.28 y 14
2006 Cincinnati Western Michigan 9.07 3 -6.07 y 15
2012 Texas Tech Minnesota 9.30 3 -6.30 y 16
2012 Vanderbilt North Carolina State 7.68 14 6.32 y 17
2007 Utah Navy 9.61 3 -6.61 y 18
2013 Pittsburgh Bowling Green State -4.34 3 7.34 n 19
2013 Oregon State Boise State 6.82 15 8.18 y 20
2010 Miami (OH) Middle Tennessee State 5.50 14 8.50 y 21
2009 Marshall Ohio -5.63 4 9.63 n 22
2014 Wisconsin Auburn 7.69 3 -10.69 n 23
2006 Boston College Navy 12.24 1 -11.24 y 24
2013 Arkansas State Ball State -8.98 3 11.98 n 25
2011 Houston Penn State 3.39 16 12.61 y 26
2013 Colorado State Washington State -10.20 3 13.20 n 27
2013 Southern California Fresno State 10.92 25 14.08 y 28
2012 Central Michigan Western Kentucky -11.36 3 14.36 n 29
2007 Fresno State Georgia Tech -3.11 12 15.11 n 30
2014 Florida East Carolina 7.19 20 15.19 y 31
2010 Pittsburgh Kentucky 1.67 17 15.33 y 32
2010 Notre Dame Miami (FL) -0.94 16 16.94 n 33
2011 Northern Illinois Arkansas State -1.59 18 19.59 n 34
2009 Florida Cincinnati 6.98 27 20.02 y 35
2007 West Virginia Oklahoma -0.52 20 20.52 n 36
2010 Northern Illinois Fresno State 0.97 23 22.03 y 37
2007 Missouri Arkansas 5.09 31 25.91 y 38
2011 Southern Methodist Pittsburgh -5.82 22 27.82 n 39
2014 Utah Colorado State 2.96 35 32.04 y 40
2012 Oklahoma State Purdue 11.28 44 32.72 y 41
2008 Tulsa Ball State -3.73 32 35.73 n 42


In the table's 42 bowl games featuring one interim coach (two games, the 2011 Fight Hunger Bowl and 2014 Armed Forces Bowl, were excluded because both coaches were interim), pure numbers devoid of any narrative context correctly predicted the result about 70 percent of the time and the games varied from their expectations by about 12 points on average.

To put some perspective on the accuracy of these interim-coached game predictions, we compared the predictions for the 42 interim-coached games to predictions for the 295 bowl games not featuring an interim coach from 2005-14.

The table below splits up each category and shows absolute values for the average expected margin of victory, actual margin of victory, difference between the two (which tells us how the games conformed to expectations), and how often The Power Rank scores predicted the winner.

Coaches N Expect Actual Diff Predict%
Interim 42 6.76 12.43 11.37 71.43%
Others 295 5.64 12.67 12.99 58.98%


Interim and non-interim coached games show remarkable similarities. For each category, there was roughly an expected touchdown difference between bowl opponents, actual victory margins were just under two touchdowns on average and the difference between predictions and outcomes was by about 12 points on average.

The only area where there was much difference between interim and other bowls was in predicted winning percentage. Despite the narrative of interim-coached games being unpredictable, interim games were actually predicted more accurately than other bowl games. But the bump in predicted winning percentage in interim bowl games shouldn't be interpreted to mean interim bowls are easier to predict. Because the interim sample is small it's likely this dichotomous variable would regress toward the mean with a bigger sample.

Interim or not, bowl games are said to be unpredictable because there's a large layoff in between games and each program is dealing with varying levels of motivation. There's research backing up the theory that bowl game results are volatile, as one study shows the spread is more accurate in predicting regular season games than bowl games. Given that bowl games can be so difficult to predict, there may be validity behind some of the narratives about rustiness and motivation. But the likelihood that a team will underperform or that there's truth to these pervasive stories is independent of whether or not there's an interim coach involved. The narratives that are frequently applied to interim-coached games could just as easily be applied to bowl games in general.

Although bowl games may be harder to predict than regular-season games, our exercise shows bowl-game predictions aren't entirely unreliable. The team that's better during the regular season usually wins the bowl game, and it doesn't appear to matter much if the coach has been in charge for one game or for 10 years. While broadcasters like to speculate how interim-coached teams will "respond" to their temporary leader, those looking to give themselves the best chance at correctly predicting results are better off ignoring this noise and sticking to their normal routine of simply predicting that the better team will usually win.  

* * *
Ross Benes is a Sports on Earth contributor who has written for Deadspin, The Wall Street Journal, Esquire and Slate. He is also working on a book about indirect relationships between sex and society. You can reach him at rossbenes@gmail.com and follow him on Twitter @RossBenes.