E. Cabral Balreira
Trinity University, USA.
Title: Using an Oracle to improve the Quality of Predictions.
Biography
Biography: E. Cabral Balreira
Abstract
There are many different approaches to rank teams in the NFL and NBA. One is able to make predictions of game outcomes and decide postseason odds. In this project, we are in interested in quantifying the accuracy and improving the quality of the predictions given by a particular ranking model. We introduce a method to address the over-confidence of game predictions that increases the quality of predictions when assessed by standard scoring metrics. In addition, we develope a method to better include homefield advantage in a ranking method. We evaluate our predictions in the past 15 years of NFL and NBA and show that our newly developed ranking method called the Oracle consistently outperformes currently available computer models in accuracy and quality of predictions.