Baseball Roster Management Dashboard: Constructing a Website for MLB Roster Visualization and Predictive Analytics Display

For this project, I constructed a website designed to display MLB roster information, edit team roster composition, and provide predictive analysis on team and individual player performance. My motivation for building this dashboard was simply a love for baseball and an interest in statistics and predictive analytics. The website is not necessarily designed for one particular person or use case –– it can be used by anyone from an avid baseball fan to a fantasy baseball player wanting to gain an extra edge. The website has three main pages, each with a different function. The first page displays historical roster data for the user’s preferred MLB team. On this screen, the user can view team and individual statistics, edit their team’s roster composition, and compare their team’s lineup and pitching rotation against an opponent’s. The second page displays individual and team projections for the upcoming season. Again, the user can edit their team’s roster and calculate a projection on an updated lineup. Finally, the third page displays fair value estimates for player salaries. These projections aim to capture what players’ average annual contract value should look like if they were signed in free agency. All estimates and predictions contained in the dashboard are generated by predictive machine learning models. Individual player predictions are produced by several long short term memory (LSTM) neural networks designed to take in season-by-season historical performance data as a time series and output future season stat lines. To predict team wins, the dashboard uses a random forest regression model trained on historical aggregate team statistics and season wins. Lastly, player fair salary estimates are produced by a neural network trained on past free agent contract values and associated historical performances.