🏎️ F1 2026 Race Prediction

A Machine Learning project that predicts Formula 1 race podium finishes for the 2026 season using historical data from 2000 to 2026.

🔴 Live App: https://f1-predictions-7.streamlit.app

---

📸 Dashboard Preview

Home	Race Predictions	Driver Analysis
2026 standings + season overview	Podium probabilities per circuit	Elo rating career history

---

🧠 How It Works

1. Data — pulls race results from 2000 to 2026 using the Jolpica/Ergast API
2. Features — engineers 12 prediction features from raw data
3. Model — trains Random Forest and XGBoost classifiers
4. Predict — outputs podium probability for each driver at any circuit

---

📊 Model Performance

Model	Accuracy
Random Forest	92.6%
XGBoost	92.1%

Trained on 10,071 rows (2000-2024), tested on 567 rows (2025-2026)

---

⚙️ Features Used

Feature	Description
grid	Starting grid position
driver_avg_finish	Weighted rolling avg finish (last 5 races)
driver_avg_points	Weighted rolling avg points (last 5 races)
driver_dnf_rate	Driver retirement rate
team_avg_points	Constructor rolling pace (last 3 years)
team_dnf_rate	Constructor reliability rate
circuit_avg_finish	Driver historical avg finish at this circuit
elo_rating	Custom Elo skill rating (resets on team change)
elo_gap_to_field	How far above/below field average Elo
team_competitiveness	Team pace relative to season average
driver_consistency	Std deviation of recent finishes
grid_penalty	Podium likelihood factor based on grid slot

---

🔧 Advanced ML Concepts Applied

• Time decay — recent races weighted higher than old ones
• Elo rating system — driver skill score updated after every race
• Team change detection — Elo partially resets when driver switches team
• Probability calibration — raw probabilities normalized for realism

---

📁 Project Structure

f1-predictions/
├── app/
│   └── app.py              # Streamlit dashboard
├── data/
│   ├── raw/                # race_results.csv (2000-2026)
│   └── processed/          # features.csv (engineered features)
├── models/
│   ├── rf_model.pkl        # trained Random Forest
│   └── xgb_model.pkl       # trained XGBoost
├── notebooks/
│   └── explore.ipynb       # EDA and data exploration
├── src/
│   ├── data_loader.py      # Jolpica API data pipeline
│   ├── features.py         # feature engineering
│   └── model.py            # model training and prediction
├── requirements.txt
└── README.md

---

🚀 Run Locally

1. Clone the repo:

git clone https://github.com/Gunjan10-droid/f1-predictions.git
cd f1-predictions

2. Create virtual environment:

python -m venv venv
venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

4. Fetch fresh data:

python src/data_loader.py

5. Build features:

python src/features.py

6. Train models:

python src/model.py

7. Run dashboard:

streamlit run app/app.py

---

📡 Data Sources

• Jolpica API (free) — race results, grid positions, standings 1950-2026
• No API key required

---

🏁 Dashboard Pages

• Home — 2026 season overview, current standings chart
• Race Predictions — select any circuit, predict podium with XGBoost or Random Forest
• Driver Analysis — Elo rating career history, 2026 race by race results
• Model Insights — feature importance charts for both models

🛠️ Tech Stack

• Python 3.9
• pandas — data manipulation
• scikit-learn — Random Forest
• XGBoost — gradient boosting
• Streamlit — dashboard
• matplotlib — visualizations
• requests — API calls

---

👨‍💻 Author

Gunjan Agrawal

• GitHub: @Gunjan10-droid

---

📜 License

MIT License — feel free to use and modify!