NCAA Tourn. "Dance Card"
CFP "FourCast"
MinV CFB Ranking
Other Sports Research
Current Ranking
Description
Performance
Author
Media Attention
Past Rankings
Acknowledgements

College Football Playoff "FourCast"    

Description

The CFP FourCast predictive formula used herein contains factors that were found to be significantly related to the CFP Selection Committee's weekly rankings during the 2014, 2015, and 2016 seasons. Each of these factors are generally known prior to the committee's announcement of its weekly rankings during the latter part of the season, meaning the model can be used for prediction purposes. As the model is still under development and subject to modification as new data becomes available, the factors included in it are also subject to change. Additionally, given that the values of some factors may become available just before or even after the committee's announcement of its weekly ranking, the ranking reported herein may reflect alternate versions of the model early in the week, and may be updated as additional factor values become available.

Factors under consideration for incorporation into the FourCast model include the traditional opinion polls from the Associated Press (media) and USA Today (coaches), as well as the rankings from the 100 or so college football ranking systems that are posted weekly on Kenneth Massey's college football ranking compilation page. Various team performance factors that can easily be independently generated are also considered, such as winning percentage, average victory margin, number of wins and losses, road record, neutral site record, conference record, non-conference record, overall strength of schedule, non-conference strength of schedule, record against top-25 teams, and conference championships. Various factors not directly associated with team performance are also considered, such as being in a Power-5 league, being the alma mater of a committee member, and being the employer of a committee member. The previous week's CFP ranking may also be considered.

Although the number of potential predictors is well over 100, the number that are included at any time in the FourCast model is much smaller. The FourCast version being used to get the current predictions has no more than 10 factors or so. (Note again that this is subject to change as additional data collection and analysis is done.)




B. Jay Coleman, Ph.D.
Richard deR. Kip Professor of Operations Management & Quantitative Methods
Department of Management | Coggin College of Business | University of North Florida | Jacksonville, FL 32224
jcoleman@unf.edu

Disclaimer