Using pitch-level data from two games I produced self-explanatory graphs and tables to summarize the players’ performance in those two games.

Bar graph insights:

Fastballs (FF) were the most frequently used pitch in both games, with similar usage across Game 1 and Game 2.

• Sliders (SL) and curveballs (CU) were more commonly used in Game 1 than in Game 2.

• Changeups (CH) and sinkers (SI) had higher usage in Game 1 compared to Game 2.

• Cutter (FC) usage was more prevalent in Game 1, while the knuckle curve (KC) was rarely used in either game.

Scatter plot insights:

• The highest spin rates in Game 1 were on Cutters (FC), where the highest spin rates in Game 2 were on Changeups (CH).

• Average Slider (SL) spin rate was higher in Game 1 compared to Game 2.

• Fastball (FF) velocities are clustered very similarly in Game 1 and Game 2.

• There was a much wider range of velocities for Changeups (CH) in Game 1, ranging from the low

80s to the 90s, whereas Game 2 had a more consistent range between 82-85 mph.

This analysis aimed to classify pitch types (Fastball, Slider, Curveball, Changeup) using various machine learning models, evaluate the performance of those models, and explore the role of different features in making accurate predictions. To achieve this, multiple methodologies and visualization techniques were employed, offering deep insights into the nuances of pitch classification. The results highlight the journey from using a Random Forest model as a baseline to leveraging XGBoost as the optimized model through hyperparameter tuning and feature engineering. This report includes the critical methodologies, challenges encountered, and advanced interpretative techniques used to achieve robust results.

This study introduces an innovative approach to quantify pitcher fatigue, leveraging the rich data provided by the Statcast system. By focusing on detailed pitch-by-pitch data, we aim to uncover how fatigue affects various pitching metrics, potentially impacting a pitcher’s performance within a game. To complement our analytical framework, we also created a Shiny application, enabling a dynamic exploration of pitcher performance data. This tool allows for interactive analysis, offering users the ability to examine and identify pitchers who may benefit from role adjustments based on their fatigue profiles and performance metrics. Integrating this application into our study enriches the decision-making process, providing a practical platform for applying our findings to real-world baseball strategy and pitcher management.