Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/Stronauta/MLB-Performance-Analytics/llms.txt

Use this file to discover all available pages before exploring further.

This guide walks you through everything you need to go from a fresh clone to a fully running classification pipeline. By the end, you will have loaded the 911-batter Statcast dataset, engineered the three target variables, trained all three Random Forest classifiers, evaluated their performance, and generated feature importance charts — all inside a single Jupyter notebook.
Python 3.9 or higher is recommended. The project badge targets 3.9+, and some Scikit-Learn and Pandas APIs used in the notebook may behave differently on older versions.

​
Steps

1

Clone the Repository

Clone the project from GitHub and move into the project directory:
git clone https://github.com/Stronauta/MLB-Performance-Analytics.git
cd MLB-Performance-Analytics
2

Set Up a Python Environment

Create an isolated environment to avoid dependency conflicts. Choose either virtualenv or conda:Using venv (built-in):
python -m venv .venv

# Activate on macOS/Linux
source .venv/bin/activate

# Activate on Windows
.venv\Scripts\activate
Using conda:
conda create -n mlb-analytics python=3.11
conda activate mlb-analytics
3

Install Dependencies

The notebook imports pandas, numpy, matplotlib, seaborn, and scikit-learn. Install them all at once, along with Jupyter:
pip install pandas numpy matplotlib seaborn scikit-learn jupyter
4

Configure the Data Path

The notebook contains a hardcoded Windows absolute path that must be updated before the notebook will run on any other machine.
The first code cell in MLBStats.ipynb sets ruta = r"D:\Python\Modulo 6\Databases\MLBDATA.csv". This path does not exist outside the original development machine. You must replace it with a path that points to Data/MLBDATA.csv inside the cloned repository.
Open the notebook and replace the ruta variable in the first code cell:
# Original (Windows path — update this)
ruta = r"D:\Python\Modulo 6\Databases\MLBDATA.csv"

# Cross-platform replacement
import os
ruta = os.path.join(os.path.dirname(os.getcwd()), "Data", "MLBDATA.csv")
# Or simply, when running from the Notebook/ directory:
ruta = "../Data/MLBDATA.csv"
The simplest option is the relative path "../Data/MLBDATA.csv", which works correctly when Jupyter is launched from the repository root and the notebook is opened from the Notebook/ subdirectory.
5

Run the Notebook

Launch Jupyter from the repository root:
jupyter notebook Notebook/MLBStats.ipynb
Once the notebook opens in your browser, run all cells in order with Kernel → Restart & Run All. The full pipeline — data loading, cleaning, feature engineering, model training, evaluation, and visualization — executes end-to-end without any additional input required.

​
What Runs in the Notebook

The notebook executes a complete machine learning pipeline from raw data to evaluated models: 1. Data Loading & Inspection The CSV is loaded into a Pandas DataFrame (df.shape confirms 911 rows × 76 columns). Basic info() and head() calls verify column types and check for nulls across all 76 Statcast features. 2. Exploratory Data Analysis Seaborn and Matplotlib visualizations explore distributions of key metrics (wOBA, launch speed, bat speed, walk rate, strikeout rate) and their relationships — including a scatterplot comparing plate discipline scores against overall offensive performance. 3. Target Variable Engineering Three classification targets are constructed directly from dataset columns:
  • Rendimiento_labels — pd.qcut on woba into Low / Medium / High thirds
  • elite_hitter — binary flag for batters at or above the 80th percentile of woba
  • clase_disciplina_home — pd.cut on a composite disciplina_en_home score (walk rate ×0.30, takes ×0.20, plate appearances ×0.10, minus strikeout rate ×0.20 and whiff rate ×0.10) into Baja / Media / Alta
4. Model Training — Three Classifiers Each target gets its own RandomForestClassifier trained on an 80/20 train/test split (stratified for the elite and discipline models):
  • Overall Performance model: RandomForestClassifier(n_estimators=300, max_depth=8, random_state=42)
  • Elite Status model: RandomForestClassifier(n_estimators=300, max_depth=8, class_weight='balanced', random_state=42)
  • Plate Discipline model: RandomForestClassifier(n_estimators=300, max_depth=8, class_weight='balanced', random_state=42) trained on the discipline-specific feature subset (bb_percent, k_percent, swing_miss_percent, swings, whiffs, takes, pa)
5. Model Evaluation Each model is evaluated with accuracy, precision, recall, and F1-score via Scikit-Learn’s classification_report. Results are printed per-class so performance on minority classes (e.g., elite batters) is clearly visible. 6. Feature Importance Visualization rf.feature_importances_ is extracted from each trained model and plotted as a horizontal bar chart, revealing which Statcast metrics most strongly drive each classification — for example, woba and xwoba for overall performance, and bb_percent and whiffs for plate discipline.

Build docs developers (and LLMs) love

Get started for freeTalk to us