This project utilizes Statcast pitch data to perform a network analysis of MLB pitchers and batters. It leverages statistical measures such as Win Probability Added (WPA) and Run Expectancy (RE) to analyze player performance using a PageRank algorithm. The goal is to identify key players and understand the dynamics of player interactions throughout the season.
- Data Import: Fetch and preprocess Statcast event data.
- Statistical Analysis: Calculate cumulative stats like WPA and RE for each player and team
- Network Analysis: Construct a network graph where nodes represent players and edges represent performance metrics.
- PageRank Calculation: Apply PageRank to assess the importance of players in the network based on their performance metrics.
These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.
Before running the project, ensure you have the following installed:
- Python 3.8+
- Pandas
- Numpy
- NetworkX
- PyBaseball
- Matplotlib
- Pickle
- Manually sourced payroll for 2023 MlB season
This script is responsible for fetching and preprocessing Statcast pitch data from the specified MLB season. Key operations include:
- Data Fetching: Using the
pybaseballlibrary, it retrieves event data between specified dates. - Preprocessing: Filters out incomplete records, normalizes player names using
unidecode, assigns players to teams, and prepares several datasets for further analysis. Additionally, all pitches which do not result in an outcome (thus strike and ball) are filtered out. - Gamescoring Goes through pitch data and assigns WPA and RE scores from each event to the corresponding players.
- Data Export: Outputs a filtered event dataset to
2023_mlb_event_data.csvand performs initial calculations of Win Probability Added (WPA) and Run Expectancy (RE) for further use. Batter and Pitcher Csv's are created to store stats for the respective players and game restults csv is created to store game results.
Builds a directed multigraph representing the interactions between pitchers and batters:
- Data Loading: Reads the preprocessed CSV files to retrieve player statistics.
- Graph Construction: Utilizes
NetworkXto create nodes for each player (pitchers and batters) and edges that represent game events. There are a maximum of two edges between each pitcher and batter, one for RE and one for WPA. These edges represent the cummulative stats for all the matchups between the respective players and the edge is directed towards the player who has the postive RE or WPA. - Node and Edge Attributes: Each node stores player stats, and edges are weighted by performance metrics (WPA, RE, scores from events).
- Graph Serialization: Saves the graph to a file
player_network.picklefor use in subsequent analysis.
Applies the PageRank algorithm to the network to identify influential players:
- Graph Loading: Loads the network graph from the pickle file.
- PageRank Calculation: Computes PageRank scores for each player using their on-field interactions and statistics like WPA and RE.
- Normalization: Adjusts the PageRank scores and recalculates them to ensure consistency across different performance metrics.
- Results Saving: Updates the graph with new PageRank scores and saves the updated graph back to a pickle file.
Handles the visualization of the network graph created and analyzed in previous steps:
- Graph Loading: Retrieves the graph from the pickle file.
- Visualization Setup: Configures node sizes and colors based on player roles and performance metrics.
- Drawing the Graph: Uses
matplotlibto draw nodes, edges, and labels. - Display Output: Renders the graph visually on screen, helping to interpret the complex relationships and player impacts visually.