By Hank Duncan
Every December for the last eight years, my dad and I compete in a college football bowl bonanza contest against the best and the brightest football minds of Floyd Memorial Hospital. While both of us peaked in 2010 with my dad reaching third place and me celebrating my first and only win of the contest, I strive to again reach the top of the standings. Now, as the contest has grown to about 100 contenders fighting for the coveted first place spot, I found myself slipping behind other competitors and back into the field. I needed a way to relive the glory days of my bowl bonanza success. As a member of a sports analytics club, the obvious solution was to use analytics!
Over the course of the past few days, I procrastinated studying for finals by creating an excel spreadsheet analyzing every college football bowl game. By breaking down each game into the FEI, S&P+, and the F/+ models, I predicted who would win each bowl game.
While I made most of my decisions based solely on analytics, the rules of this contest caused me to switch a few of my picks. As seen at the bottom of the article, the rules state that, if you pick an underdog to win the game, the game value put on that game doubles. So, it essentially gives an incentive to play it risky and pick underdogs.
After choosing who would win each game, I needed to put a game value for each game. The game value represents how confident I am in that decision. They range from 1 to 42 (there are 41 bowl games plus the championship), and I may use each number only once. However, as I previously mentioned, if an upset wins, that game value doubles.
To decide which value to put on each game, I calculated the expected points from each game. In order to find out a rough estimate of the expected points from a game, I multiplied the S&P+ win probability by each game value. For example, if I pick Houston to beat San Diego St. and give them a full 42 points, I would multiply 58.3% (win probability) by 42. So, Houston’s expected point total is 24.5. However, if I pick UTSA (underdog) to beat New Mexico and give them a full 42 points, I would multiply 33.8% (win probability) by 42 and then double it because it is an upset. So, UTSA’s expected point total it 28.4.
Through this analysis and a bit of my own judgement, these are my college football bowl picks for this year’s bowl bonanza contest.
|Bowl Game||Favorite||Underdog||My Choice||Game Value||Expected Points|
|Celebration||Grambling St||NC Central||Grambling St||20||15|
|New Mexico||New Mexico||UT San Ant||UT San Ant||31||20.1|
|Las Vegas||Houston||San Diego St.||Houston||11||6.4|
|Camelia||Toledo||App St.||App St.||35||35.4|
|Cure||Central Florida||Arkansas St.||Central Florida||4||2.5|
|New Orleans||So. Miss.||Louis. Laf.||So. Miss.||13||8.7|
|Miami Beach||Tulsa||Central Mich||Tulsa||21||15|
|Boca Raton||Western Kent||Memphis||Western Kent||7||4.9|
|Idaho Potato||Colorado St.||Idaho||Colorado St.||29||24.7|
|Bahamas||Old Dominion||Eastern Mich||Eastern Mich||27||21.4|
|Armed Forces||Navy||Louis. Tech||Louis. Tech||39||48.3|
|Hawaii||Hawaii||Mid Tennessee||Mid Tennessee||42||52.2|
|St. Petersburg||Miss. St.||Miami Oh.||Miss. St.||2||1.3|
|Quick Lane||Maryland||Boston Coll.||Boston Coll.||38||37.8|
|Independence||NC State||Vanderbilt||NC State||12||9|
|Heart of Dallas||Army||North Texas||Army||14||9.3|
|Holiday||Wash St.||Minnesota||Wash St.||1||.5|
|Cactus||Boise St.||Baylor||Boise St.||28||23.6|
|Russell Athletic||Miami||West Virginia||Miami||15||11|
|Texas||Texas A&M||Kansas St.||Kansas St.||26||20.6|
|Birmingham||South Florida||South Carolina||South Florida||23||18.3|
|Belk||Virginia Tech||Arkansas||Virginia Tech||16||10.6|
|Sun||Stanford||North Carolina||North Carolina||36||35.6|
|Arizona||Air Force||South Bama||Air Force||19||13.1|
|Taxslayer||Georgia Tech||Kentucky||Georgia Tech||11||6.8|
|Fiesta||Ohio St.||Clemson||Ohio St.||4||2.4|
|Rose||USC||Penn St.||Penn St.||33||27.8|
|CFP National Championship||Alabama||Ohio St.||Ohio St.||32||23.5|
Based solely off of the S&P+ win probability, my total expected points for the bowl bonanza contest is 803.5 points.
However, I wanted to run simulations to test these probabilities. I created a data table of 1,000 simulations of each bowl game. Using the win probabilities, I used various IF and RAND functions to create a formula that simulates game and returns the number of points I would receive if the team I picked wins or loses. In essence, it represents 1,000 different versions of the 2016-2017 bowl season.
From these simulations, I took the 1,000 different total points expected and found the average. The total points expected for the bowl contest based on binomial simulations is 824.3 points with a standard deviation of 150.5. To give a better representation of the spread and distribution of each total points expected, the chart below depicts a few percentiles and its respective score.
Binomial Simulation Percentiles
As the 2016-2017 bowl season kicks off, I wish all of you well in your bowl contests. Just remind yourself that, before you pick the team you love or pick against the team you hate, the glory at the end of the contest feels much better than that dirty feeling of going against your gut feeling. If I end up winning this year, I will feel much better about picking Ohio State and against Indiana. Good luck and happy bowl season!