Monday, January 17, 2011

Using the Point Spread to Analyze Bowl Performance

If you spent any time this bowl season watching ESPN after New Year's Day, you no doubt heard the talking heads going on and on about how 'awful' the Big 10 was. They loved to site and ridicule in particular the league's 0-5 mark on January 1st. However, what they failed to mention was that all 5 Big 10 teams were underdogs in those games, with Northwestern (9 and a half point dogs to Texas Tech) and Michigan State (10 point dogs to Alabama) coming in as prohibitive underdogs. In order to come up with a more sophisticated way to look at how a conference (or team) performs in the bowl season, it would help to know the probability they had of winning their bowl game. One way to estimate win probability is to use the point spread or Vegas line. While the point spread is to some degree influenced by public opinion (in the form of line shifts), it is for the most part an unbiased rating system. In order to determine how likely a team with a given point spread was to win a particular game, I looked at all games played between IA teams from 2005-2009 where there was a point spread of at least one point (i.e. the game was not a pick em' or had a half point favorite) and determined who won the actual game (covering the spread was not important in this exercise). I then (somewhat) arbitrarily divided the results into difference ranges. They are listed in the table below.For the most part, the numbers are pretty intuitive. When a team is favored by less than a field goal, the game result is usually a coin flip, with the favorite prevailing just over 50% of the time. When the margin increases to at least a field goal, but less than a touchdown, the favorite wins nearly 60% of the time. You can peruse the rest of the table for yourself, and see if you can guess the 2 teams that lost as favorites of 30 points or more (the answer is at the bottom of this post). I'll give you a hint, they both happened in 2007 and occurred within 2 weeks of each other.

It is pretty simple to use these probabilities to estimate how many games a conference 'should have' won in the 2010 bowl season. I'll go through the numbers for the Big 10 and then show you how each conference performed. The Big 10 had 8 teams participating in the 2010 bowl season. Ohio State was a 3.5 point favorite over Arkansas, meaning they had roughly a .599 chance of winning the game based on similar point spreads over the last 5 years. Northwestern was 9.5 point dog to Texas Tech meaning they had about a .254 chance of winning (that is 1 minus the chance of a 9.5 point favorite winning or 1-.746). Michigan State was a 10 point dog to Alabama, giving them a .254 chance of winning. Penn State was a 7 point dog to Florida, giving them a .295 chance of winning. Michigan was a 5 point dog to Mississippi State, giving them a .401 chance of winning. Wisconsin was a 3 point dog to TCU, giving them a .401 chance of winning. Iowa was a 3 point dog to Missouri, giving them a .401 chance of winning. And finally, Illinois was a 1.5 dog to Baylor, giving them a .492 chance of winning. Add all those win probabilities up and we come up with 3.097 expected wins for the Big 10 based on the point spread. The Big 10 actually won 3 bowl games (Ohio State, Iowa, and Illinois), meaning they underperformed their expected win total slightly. Here is how every conference performed in the 2010 bowl season.The Big 10 was a respectable 7th among all 12 conferences. They were 6th of 11 if we remove the independents which is not really a conference and only includes 3 teams (Army, Navy, Notre Dame). The conference that performed the worst by far was the Big 12, and yet there was hardly a whimper of their poor performance on ESPN. The main culprit for the Big 12's bowl woes was Nebraska. The Cornhuskers entered the bowl season as the 3rd largest favorite (after Oklahoma and Boise), but were upset by the 14 point dog Washington Huskies. In a nice twist of fate, the Huskers are leaving the conference that performed the worst in the 2010 bowl season and joining the conference that was perceived to have performed the worst.

Southern Cal to Stanford (as 41 point favorites) and Louisville to Syracuse (as 37 point favorites) are the 2 teams to lose as 30+ point favorites.

No comments: