NCAA Power Rankings – Final Rankings Using Cross-Validation

As the final week of the regular season played out in the NCAA Division I FBS, we witnessed Michigan State power to a 16-13 win over Iowa to win the Big Ten as Clemson managed to weather a comeback attempt from North Carolina for a 45-37 win en route to winning the ACC. In our latest rankings model, we had  a couple teams that were penalized harshly for close losses to inferior teams, while other teams benefitted from losses to superior teams. To counter-act this from occurring, we perform a leave-one-out cross-validation to create our rankings.

There were 713 total FBS games in the NCAA this year. The cross-validation created 713 rankings where one game has been eliminated from the season. This method will take into account the best and worst games for each team. Once we have the 713 total rankings, we can perform any of the following methods: maximum ranking, minimum ranking, median ranking, average ranking, or another calculation altogether. For our situation, we consider removing only the worst game for each team and build the maximum rankings. In doing this, we obtain the following NCAA FBS Championship tournament:

1. Clemson Tigers (13 – 0): ACC Champion vs 4. Oklahoma Sooners (11 – 1): Big 12 Champion

2. Michigan St. Spartans (12 – 1): Big Ten Champion vs. 3. Alabama Crimson Tide (12 – 1): SEC Champion

Taking the maximum scores, we normalize the ranking by dividing Clemson’s ranking across every other team. We obtain the remaining rankings for 5 through 128, helping us determine who should most likely be in the remaining bowl games.

RANKINGS: Final Regular Season Edition

Screen Shot 2015-12-06 at 11.07.05 AM

Final NCAA Division I FBS Rankings based on Leave-One-Out Maximum Cross-Validation: 1 through 35.

Screen Shot 2015-12-06 at 10.47.35 AM

Final NCAA Division I FBS Rankings based on Leave-One-Out Maximum Cross-Validation: 36 through 70.


Screen Shot 2015-12-06 at 10.48.21 AM

Final NCAA Division I FBS Rankings based on Leave-One-Out Maximum Cross-Validation: 71 through 105.

Screen Shot 2015-12-06 at 10.48.44 AM

Final NCAA Division I FBS Rankings based on Leave-One-Out Maximum Cross-Validation: 106 through 128.


Selecting teams for bowl games is a difficult task as many of the teams are not matched up based on their actual game play, but rather a set of loosely defined rules and determination of which games will seem to make the best match-ups. Dues to this, we may find teams are in games above their level, or playing well below their level. In our predictions of match-ups, we apply the loose rule set to the rankings above.

To be bowl-eligible, a team must have a winning record (including at most one FCS win), have won their conference, or have a pre-determined match-up (BYU, Navy, and Army). This year, 72 of the 128 teams are bowl-eligible for the 40 NCAA FBS bowl games. This means that 8 ineligible teams are selected to fill the remaining void. So teams like Texas, Auburn, Arizona, Arizona State, Washington, Indiana, Kansas State, Connecticut, and Virginia Tech all have a significant chance to get in.


Rose Bowl: Stanford (11-2) Pac-12 Champ vs. Iowa (12-1)

 Sugar Bowl: Oklahoma State (10-2) vs. Florida (10-3)

Peach Bowl: Ohio State (11-1) vs.Notre Dame (10-2)

Fiesta Bowl: Texas Christian (10-2) vs. Houston (12-1)

In determining the other bowl games, we had a difficulty as Navy is listed to play in two bowl games: Military Bowl and Poinsettia Bowl. Since the definitions of bowl games are not clearly detailed, making the other 34 bowl game picks are difficult. We have a predicted match-up, but close to 20 will be way off since there are situations such as no #9 SEC team eligible with no clearly defined back-up team.





Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.