Much has been written about the 2013 Iron Bowl, the game that won an ESPY for “Game of the Year” thanks to the play that won an ESPY for “Play of the Year.” Chris Davis’ game-winning 109-yard field goal return for a touchdown ended Alabama’s hopes for a third straight National Championship, and completed one of the biggest single-season turnarounds in NCAA history. But the game’s outcome, like every game’s, was a function of numerous big plays, mistakes, swings of momentum and luck. This series will investigate the highs and lows of each game, using the latest in trends in big data and analytics.
The data used in this model come from a 1.3 million play-by-play dataset covering every play in college football since the 2005 season. This data was used to create a series of statistical models that explain how a team tends to perform in any situation. The model discussed in this article predicts a team’s probability (odds) of winning the game. This in-game probability model “predicts” how likely a team is to win the game at any time based not only on what has already occurred in the game, but what the model expects will occur in the future. Most importantly, this model takes into account the relative strengths and habits of teams using their season-to-date tendencies. The model also leverages other important information like the Las Vegas line and total. Integrating all of this information into a statistical model allows me to provide an accurate picture of the game’s outcome at any point in the game. By charting these predictions over time, we can observe how scores, long drives, big stops, and, of course, missed field goals all impact the game’s outcome.
Alabama at Auburn, Nov. 30, 2013:
The table below charts Auburn’s probability of winning at each moment in the game (more accurately, before each snap). I will discuss some of the bigger swings in the chart and do my best to explain how the model arrived at some of its predictions. (Before proceeding, a caveat: the model’s predictions are developed from just nine years’ worth of data. This implies that extremely rare occurrences are not likely to be well-represented in data that covers just nine years in a sport that has been played for well over a century. As such, the model will tend to underestimate the true likelihood of rare events if they happen to be underrepresented in the data available.)
- The game opens with a 10-point favorite Alabama electing to receive. With no other information to go on, the model leverages the historical winning percentages for 10-point home underdogs and assigns Auburn a 26% chance of winning. It is important to understand that the value embedded in the Las Vegas line persists through the entire game. Favorites tend to win games when they are winning, tied, and even losing. This explains in part why when Auburn scored its first touchdown and went up 7-0 that the Tigers “only” had a 50% chance of winning, despite the early lead.
- Alabama’s first drive concluded with a missed field goal, at which point Auburn had a 36% chance of winning.
- About 8 minutes into the game, Auburn takes over at their own 34 and drives about 20 yards before Nick Marshall runs 45 yards for a touchdown. A more measured drive with no big plays would have provided a straighter line between the drive’s beginning and end. However, the 45-yard touchdown creates a prominent spike denoting the relative unexpectedness of the play and its importance for Auburn’s chances of winning, which were about 51% after the score.
- After the team’s traded punts and field position to close out the first quarter, Alabama completed a long 67-yard drive with a touchdown, which lowered Auburn’s chances of winning from 51% at the beginning of the drive to 36% at the end of the drive. Note that with the game now tied 7-7 and 41 minutes of game time remaining, Auburn’s chances of winning have increased from the 26% at kickoff.
- Auburn’s fumble on its next drive and Alabama’s quick scoring drive give the Tide a 7-point lead and lower Auburn’s chances to 21%.
- Auburn goes backward on its next drive and Alabama again scores quickly to make their lead 21-7. The model is observing a game where the road team is a significant favorite and is already covering the line with four minutes left in the 1st half. Auburn has just 100 yards of offense and has allowed three consecutive touchdowns. Auburn’s probability of winning is now just 5%.
- A series of nothing but rushing plays results in an Auburn score and decreases the Alabama lead to just 7. With the Tide running out the clock to end the half, Auburn’s chances of winning rise to 20%.
- The Tigers’ opening touchdown drive is shown as a relatively straight line due to the lack of explosive plays or losses. The touchdown ties the game at 21-21, but Auburn’s chances of winning remain relatively small at 23%.
- As the tie game persists and Alabama misses another field goal, Auburn’s chances rise to 30% as Alabama takes possession on its own 1-yard line.
- Just 14 seconds and a 99-yard pass play later, Alabama is up 28-21 and Auburn’s chances fall to just 11%.
- Auburn is stopped on 4th down on its next drive, but manages to do the same to Alabama on their ensuing drive. After being forced to punt on 4th and 18 from their own 5-yard line, Auburn blocks an Alabama field goal and the Tiger’s chances of winning are about 15% with less than 3 minutes left in the game.
- After a six straight runs by Tre Mason, Nick Marshall throws a 39-yard touchdown pass to Sammie Coates, and the Tigers tie the game 28-28. With less than a minute remaining in the 4th quarter, the score increases the Tiger’s probability of winning to 35%. This probability reflects the fact that Alabama has enough time remaining to get into field goal position and win the game. (The model, however, does not currently use current game performance as an explanatory factor. If it did, Auburn’s odds of winning would have been slightly greater given the Tide’s 0-3 field goal conversion to that point.)
- With a fortuitous if not debatable one second remaining on the clock, Alabama freshman Adam Griffith attempted at 57-yard field goal to win the game. The odds of a professional kicker converting a 57-yard field goal are around 40%, the odds of a collegiate freshman doing the same thing are significantly lower. Yet, Auburn’s odds of winning the game even with overtime a strong possibility were still relatively low, about 35%. This percentage reflects the likely outcome in overtime when playing the #1 team in the country as a 10-point underdog.
But as we all know, the game did not go into overtime. Chris Davis’ once-in-a-lifetime “Kick Six” capped Auburn’s incredible comeback to defeat the undefeated and top-ranked Crimson Tide. Just how incredible was that comeback? Now you know.
UPDATE: Twitter user @BIG_GAME_BUS gave the graph an OMG update:
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 degree in economics from UNC-Chapel Hill and has written numerous academic publications on college football, the NFL and gambling markets. He welcomes any and all feedback and can be reached at firstname.lastname@example.org. You can find his previous TWER columns here.
* Alabama fan Joe Scarborough cracks joke about Auburn paying players on MSNBC’s ‘Morning Joe’ because he’s still mad about the Iron Bowl
* Former MTV “Real World” call Auburn “God’s country.”
* Hota Kotb describes watching the game with her family, says ‘red shirt freshman’ Chris Davis made her brother cry
* ‘Greatest game ever’: Auburn makes a football fan of Barbara Walters