Predicting MLB Hall of Fame Induction

INFO 523 - Summer 2025 - Final Project

Austin Cortopassi, David Pelley, Nathan Harville

Abstract & Motivation

• Hall of Fame selection is one of the most prestigious yet subjective honors in Major League Baseball.

Project Goal

• Build a machine learning model to predict Hall of Fame induction using career-level statistics and achievements.

Data Source

• Lahman Baseball Database — player records spanning 100+ years

Approach

• Data cleaning & feature engineering
  
• Model training with classification algorithms

Outcome - Evaluate prediction accuracy
- Identify key career metrics that influence Hall of Fame voting outcomes

Exploratory Data Analysis

• Visualized metric distributions and adjusted data for modeling
  
• Built correlation matrices to guide feature engineering
  
• Split models:
  • Pitchers (ERA, Wins, Strikeouts)
    
  • Position Players (AVG, HR, Hits)

Feature Engineering - Batting Avg, Slugging %, On-Base %
- WHIP, K/BB

Modeling Approach

• Random Forest → interpretable, feature importance
  
• XGBoost → handles complex nonlinear patterns
  
• GaussianNB → simple baseline model

Hyperparameter Tuning - RandomizedSearchCV

Preparation - Train/test split
- Scaling + PCA experiments

Modeling Evaluation

Metrics Used - Accuracy → overall correctness
- Precision & Recall → false positives/negatives
- F1 Score → balance of precision & recall

Random Forest Results

XGBoost Results

Naive Bayes Results

Model Accuracy Results

Model Accuracy

Conclusion

• Career stats can predict Hall of Fame induction with high accuracy
  
• Models capture patterns but miss subjective factors (awards, reputation, historical context)
  
• Next steps: add advanced metrics & era adjustments for better accuracy
  
• Data analytics offers valuable insight into what drives Hall of Fame voting