Prediction of the 1996 super bowl an application of the AHP with feedback

Thomas Saaty
Joseph M. Katz Graduate School of Business
University of Pittsburgh
United States
D.S. Turner
Atmospheric Science and Technology Directorate
Environment Canada

Publication date: Jul, 1996

Journal: Proceedings of the Fourth International Symposium on the Analytic Hierarchy Process (ISAHP’96)
- Pages: 12-15

Publisher: Creative Decisions Foundation

Abstract: This paper develops a refined benefiticost model using the supermatrix approach of the AHP to predict the outcome of the playoffs and the Super Bowl of the 1995-1996 season. Introduction In October 1995, we undertook the task of predicting (Sooty and Vargas, 1991)' the outcome of the 1996 Super Bowl. At that time the Miami Dolphins, wile:, had the best statistical record of all the teams in the NFL, appeared to be destined to win the Super Bowl. Soon after, their fortunes went on the decline, but we continued predicting other likely winners using the same model. The variations in our forecasting the winner of the Super Bowl from week to week, made us ask how we could change the model to do better. Soon we learned two lessons. The first lesson was that our model developed a priority for each team from their statistical past performance and from our interpretation of their abilities, independently from who their opponents were in the past or might be in the future. In addition, we forecast the Super Bowl winner at any time, without relating it to the week by week 1;,vinners. In other words, we looked at the end without considering how we might predict the progress of the outcome from week to week Needless to say, our predictions were in error which became more appM-ent when we applied the same approach to predict the week by week outcomes of the playoffs. We asked ourselves where we could modify our understanding to better capture the realities of the situation and make correct predictions.

Keywords: Prediction, AHP, AHP with feedback, Analytic Hierarchy Process