Abstract

NCAA Division I Men’s Basketball at-large bids are widely studied every year to determine if the NCAA committees decisions are representative of what teams should be in the tournament. Different mathematical predictive methods have been used in order to determine if these at-large bids may be predicted. We sought out to determine if these same methods could be applied to NCAA Division I Softball in an attempt to predict at-large bids over the 2010-2017 seasons. We determined using a Rating Percentage Index only method, where the top 32 ranked teams in RPI who did not receive an automatic bid within every season resulted in the most accurate method with 93% accuracy.

We took game logs for every team in Division I which included: teams, opponents, scores, and location. We then created 20 possible predictor variables that could be used for various models. Next, we used a stepwise logistic regression model in order to narrow the predictor list to include the most important variables. Along with this list of predictors, we chose four more known to influence the committee allotting at-large bids. In MatLab we built two logistic regression models and two decision tree models using the stepwise logistic regression variables and the other four variables separately. Finally, we compared these models to evaluate their predictive accuracy. Figures and tables illustrating results and importance of variables for each model are found within the text.

Advisor

Pasteur, R. Drew

Department

Mathematics

Disciplines

Mathematics | Physical Sciences and Mathematics

Publication Date

2018

Degree Granted

Bachelor of Arts

Document Type

Senior Independent Study Thesis

Download

Share

COinS
 

© Copyright 2018 Madelynn A. Chase