# Are underdogs winning more bowl games?

If you were to have cast a Texas Tech – Central Florida – Oklahoma money line parlay (i.e., betting each of those teams to win) at the beginning of the bowl season you’d be expecting about a 100:1 payout, as each of the three double-digit underdogs pulled huge upsets.

Are these upsets to be expected?  Or is it unique to the 2013 bowl season?

My hypothesis (h/t to @mandelSNY and @jeffreytinker for the idea) is that bowl games are more random than regular season games, due primarily in part to (i) varying motivation levels and (ii) an extra long layoff between the end of the regular season and each bowl game.  If this randomness exists – and that’s a big if – I was mostly curious if one could use take advantage of the frequent upsets by betting on the underdogs to win each game.

To help answer this question, I started by extracting spread and score data from the Sunshine Forecast website for every bowl game since 1992. Next, I used SBR’s money line calculator to try and determine the payout taking each of the underdogs to win. For example, if a team was a 3 point underdog, betting \$100 on that team to win would pay approximately \$125. A heavier underdog, for example Oklahoma (14.5 point underdog, payout \$620), would yield a larger payout for their less than likely wins.

Here’s a (somewhat lazy) scatter plot of game spreads and finishes for all the games in this data set. The positive correlation is what we’d expect: as the home team is favored by more points, they are more often than not winning by more points.

Next, for each year, I pretended to bet \$100 on each underdog to strait up win. In 2013, with several upsets, this would have obviously been a profitable scenario, as doing so would have put you ahead roughly \$1500, as shown in the table below:

While profitable in 2013, betting each underdog to win has not always been a great idea. Just two years ago, for example, doing so would’ve been quite costly (loss of \$1270).

Of course, all underdogs are not equal: what if you were to separate this strategy by the type of underdog each team was? To do this, I stratified by the point spread in each game, using traditional football cutoffs (like 3, 7, 10, 14, etc).

Here’s a similar table using all of the games in the sample, separated by point spread interval.

Interesting. Going back roughly 20 years, if you had bet \$100 on each underdog of 10 points or more to win straight up, you’d certainly have made a nice profit (helped, in part, by the 2013 results). Sadly, there are no remaining double digit dogs left in the 2013 postseason, so we’ll have to wait until next year to attempt this strategy for real.