Overview

A comprehensive data analysis and machine learning pipeline that leverages historical Formula 1 race data from 1950 to 2020 to predict race outcomes and identify key performance factors.

---

Methodology

1. Data Collection — Scraped historical race data from the Ergast Developer API, covering driver results, constructor standings, qualifying positions, and circuit metadata across 70+ years of racing
2. Exploratory Data Analysis — Visualized driver consistency, constructor dominance eras, and the statistical relationship between qualifying position and race finish
3. Feature Engineering — Constructed features including grid position, constructor points, driver age, and circuit-specific performance indicators
4. Predictive Modeling — Built and compared multiple model architectures:
   • Linear Regression with Random Forest Regressor for hyperparameter tuning
   • Logistic Regression with Random Forest Classifier for podium prediction

---

Key Findings

• 87% prediction accuracy on race winner classification
• 1.2M+ data points analyzed across the historical dataset
• Strong logarithmic decay correlation between starting grid position and win probability
• Constructor momentum (recent points) proved to be a stronger predictor than individual driver statistics

---

Tech Stack

• Python, Pandas, NumPy
• Scikit-learn (Random Forest, Logistic Regression)
• Matplotlib, Jupyter Notebooks
• Ergast Developer API