This thesis presents an advanced machine learning approach to predicting National Football League (NFL) player performance metrics with a particular emphasis on supporting real-time pricing of Daily Fantasy Sports (DFS) Pick'em entries. Utilizing comprehensive play-by-play data from the 2023-2024 NFL regular season provided by WagerWire, the research evaluates the predictive accuracy of three prominent gradient boosting algorithms-XGBoost, LightGBM, and CatBoost. These models are designed to forecast key player statistics during live games, enabling dynamic pricing adjustments in WagerWire's secondary marketplace. The primary goal of these predictive models is to support WagerWire's FantasyWire marketplace by providing accurate, real-time pricing for DFS Pick'em entries, enhancing liquidity and user confidence. Leveraging probabilistic market data and advanced machine learning, the models demonstrate robust predictive accuracy across key player metrics, significantly improving the transparency and efficiency of marketplace transactions. A comprehensive evaluation demonstrates the model-specific strengths of each algorithm, highlighting XGBoost's consistent performance across multiple player statistics, CatBoost's precision in predicting complex outcomes such as Passing Yards, and LightGBM's efficiency and balanced accuracy. Overall, this research significantly contributes to enhancing the operational efficiency and pricing accuracy of WagerWire's secondary marketplace. Future research directions include integrating additional real-time data streams, further improving model interpretability, and extending predictive modeling frameworks to encompass diverse sports contexts.