From the start, DeepBall has been dedicated to providing you with high-quality data and predictive models to suit your baseball desires and needs. These models take time, energy, and experience to develop and perfect. At the forefront of this movement is DeepBall Labs, where we apply state-of-the-art machine learning techniques to difficult problems in baseball. In doing so, we hope to unite the ever-expanding field of machine learning with the storied past and exciting horizon of baseball analytics.
The idea behind DeepBall was conceived during the writing of Daniel Calzada's Master's thesis, centered around generating preseason batter predictions using recurrent neural networks. Since then, the work has been distributed to the Society for American Baseball Research (SABR) and Retrosheet. Among other things, we introduced our Gaussian uncertainty modeling technique where we can quantify general uncertainty in our own predictions, leading to a whole new dimension of baseball predictive modeling. The full academic work is freely available at these links, but for those who are looking for an abbreviated and less technical version, we have a summary version as well.
Already, we have more ideas in the research pipeline than we know what to do with, but we want to hear even ideas more from you! Are you a fan who loves analytics and has a great idea for a prediction or modeling approach? Are you a member of a team front office with a need for some deep data analysis? Either way, please get in touch with us with what you'd like to see happen, and we'll look into turning your need or dream into reality.
Check back here often to stay up-to-date on our research!