Determining Hall of Fame Status for Major League Baseball Using an Artificial Neural Network

Abstract

Election into Major League Baseball's (MLB) National Hall of Fame (HOF) often sparks debate among the fans, media, players, managers, and other members in the baseball community. Since the HOF members must be elected by a committee of baseball sportswriters and other entities, the prediction of a player's inclusion in the HOF is not trivial to model. There has been a lack of research in predicting HOF status based on a player's career statistics. Many models that were found in a literature search use linear models, which do not provide robust solutions for classification prediction in complex non-linear datasets. The multitude of possible combinations of career statistics is better suited for a non-linear model, like artificial neural networks (ANN). The objective of this research is to create an ANN model which can be used to predict HOF status for MLB players based on their career offensive and defensive statistics as well as the number of career end of the season awards. This research is limited to investigating players who are not pitchers. Another objective of this report is to give the audience of this particular journal an overview of ANNs.

Predicting Win-Loss outcomes in MLB regular season games – A comparative study using data mining methods

“…Artificial Neural Networks (ANNs) have been widely used for sport predictions (Aslan & Inceoglu, 2007;Edelmann-Nusser et al, 2002;Young, Holland, & Weckman, 2008). The method is a product of early artificial intelligence work aimed at modeling the inner workings of the human brain, as a way of creating intelligent systems.…”

Non-invasive detection of the anaerobic threshold by a neural network model of the heart rate—work rate relationship

“…The hidden layer neurons of the MLP have tangential sigmoid transfer functions and the output layer has a linear transfer function. The training of the network was carried out by the Levenberg-Marquardt backpropagation algorithm, which generally produces the best results [11].…”

“…In a recent study an ANN modelling approach used for estimating energy expenditure with different dynamic inputs (accelerometry, HR above resting, and electromyography) [21]. The most common form of using ANN in sports research models is to predict the outcome of a sporting event [11]. One predictive system forecasted the outcomes of rugby matches in the 2003 Rugby Union World Cup [22] at a high rate of success.…”

Using Tree Ensembles to Analyze National Baseball Hall of Fame Voting Patterns: An Application to Discrimination in BBWAA Voting

“…The most recent and sophisticated prediction analyses include Freiman (2010), Smith and Downey (2009) and Young et al (2008), in which the authors use variants of Artificial Neural Network (ANN) and Machine Learning methods to predict future Hall of Fame induction. Young et al (2008) pioneered the use of the ANN in exploring Hall of Fame decision boundaries, employing a complex set of input variables to train the network. The strength of the analysis is its extremely low error rate and the ability to conduct sensitivity analysis: a way to find the most important input variables in the network.…”

“…However, these models restrict the decision of the BBWAA to be linear using a logistic (or classical OLS) regression model. As noted by Young et al (2008), given the diverse inputs that BBWAA voters likely take account for when making their decisions, a linear model could be limited in its ability to predict Hall of Fame induction. Previous economic models of voting behavior have focused particularly on discrimination against African American and Hispanic players.…”