Title:
NFL & NCAA Football Prediction using Artificial Neural Networks
Authors:
Blaikie, Andrew D.; Abud, Gabriel J.; David, John A.; Pasteur, R. Drew
Publisher:
Denison University
DATE ISSUED:
19-Nov-2011
PERMANENT LINK:
http://hdl.handle.net/2374.DEN/3930; http://hdl.handle.net/2374
Type:
Article
Language:
en_US
Description:
The modeling of predicting the outcome of football games is very challenging. Even the best predictive models have an average absolute error of 10-12 points per game in the NFL and 12-14 points in Bowl Championship college football. Artificial neural networks were used to create models to predict the outcome of football games for both the NFL and college football. The NFL model was a continuation of the model in [1] and the college football model was new. Data analysis was done to identify the most predictive statistics, which were later used in the model. The model used was based purely on statistics and used a committee of machines approach for greater consistency. Many models were compared to determine which was the most accurate. It was found that the college football model performed poorly when compared to the NFL model. We discuss reasons for these results and procedure to overcome the challenges. Afterwards, the models were examined using derivative analysis. The results of the research showed that the NFL model consistently was in the top half compared to other prediction experts, while the college football model tended to be closer to the middle of these rankings.
Appears in Collections:
Proceedings: 2011 Midstates Conference for Undergraduate Research in Computer Science and Mathematics

Full metadata record

DC FieldValue Language
dc.contributor.authorBlaikie, Andrew D.en
dc.contributor.authorAbud, Gabriel J.en
dc.contributor.authorDavid, John A.en
dc.contributor.authorPasteur, R. Drewen
dc.coverage.spatialUSA - Ohio - Licking - Granvilleen_US
dc.date.accessioned2012-10-11T20:42:35Zen
dc.date.accessioned2013-12-18T22:13:36Z-
dc.date.available2012-10-11T20:42:35Zen
dc.date.available2013-12-18T22:13:36Z-
dc.date.created2011-11-19en
dc.date.issued2011-11-19en
dc.identifier.urihttp://hdl.handle.net/2374.DEN/3930en
dc.identifier.urihttp://hdl.handle.net/2374-
dc.descriptionThe modeling of predicting the outcome of football games is very challenging. Even the best predictive models have an average absolute error of 10-12 points per game in the NFL and 12-14 points in Bowl Championship college football. Artificial neural networks were used to create models to predict the outcome of football games for both the NFL and college football. The NFL model was a continuation of the model in [1] and the college football model was new. Data analysis was done to identify the most predictive statistics, which were later used in the model. The model used was based purely on statistics and used a committee of machines approach for greater consistency. Many models were compared to determine which was the most accurate. It was found that the college football model performed poorly when compared to the NFL model. We discuss reasons for these results and procedure to overcome the challenges. Afterwards, the models were examined using derivative analysis. The results of the research showed that the NFL model consistently was in the top half compared to other prediction experts, while the college football model tended to be closer to the middle of these rankings.en_US
dc.language.isoen_USen_US
dc.publisherDenison Universityen_US
dc.relation.ispartofProceedings of the 2011 Midstates Conference on Undergraduate Research in Computer Science and Mathematicsen_US
dc.titleNFL & NCAA Football Prediction using Artificial Neural Networksen_US
dc.typeArticleen_US
dc.contributor.institutionDenison Universityen_US
dc.contributor.repositoryDenison DRCen_US
dc.publisher.digitalDenison Universityen_US
All Items in The Five Colleges of Ohio Digital Repository are protected by copyright, with all rights reserved, unless otherwise indicated.