Evaluating Baseball Statistics by Predicting Playoff Teams

Abstract

In this paper, we explore how different baseball statistics correlate to an entry to the playoffs. We use LogisticRegression and XGBoost to evaluate if a baseball statistic has a high correlation with whether or not a team makes the playoffs. We set up three models: the Moneyball model (which uses moneyball statistics), the All Stats Model (which uses moneyball statistics and additional common statistics), and the XGBoost model (which uses the same dataset of the All Stats model, but the structure of the model is different). We compare these models, evaluating their accuracy, variable coefficients, and confusion matrix. From these tests, we find that the Moneyball Model has similar accuracies to the All Stats Model, revealing that moneyball statistics are still a relevant and accurate way to predict if a team makes the playoffs. The variable coefficients test highlights that moneyball statistics have the highest importance in the model's ability to predict if a team makes the playoffs. While the tests provide a foundation for the evaluation of moneyball and common baseball statistics, there remains future opportunities to use different models and a larger dataset.