ScoreCast – Cricket Score Prediction

A Flask-based web application that predicts the final T20 cricket score of a batting team given the current match situation, built with Python, scikit-learn, XGBoost, and deployed via Flask.

Demo

---

Features

• Select batting team, bowling team, and venue (city) via dropdowns.
• Input current score, overs completed (e.g., 18.2 for 18 overs and 2 balls), wickets fallen, and runs scored in the last 5 overs.
• Dynamic country flags update when teams are selected.
• Real-time prediction of final team score using a trained XGBoost regression model.

---

Technology Stack

• Backend: Python 3.x, Flask
• Data Processing & ML: Pandas, NumPy, scikit-learn, XGBoost
• Templating: Jinja2 (Flask)
• Model Persistence: pickle
• Frontend: HTML, CSS, JavaScript

---

Dataset

• Source: Collected ball-by-ball data from T20 international matches (post-2021 World Cup up to 2024 venues).
• File: T20_iore_info.csv
• Columns:
  • match_id: Unique identifier
  • batting_team, bowling_team: Team names
  • ball: Over.ball format (e.g., 5.3)
  • runs: Runs scored on that delivery
  • player_dismissed: Dismissal info (if any)
  • city: Venue city (missing for some rows)

---

Feature Engineering

1. City Imputation & Filtering: Fill missing city values by extracting from venue, then keep cities with ≥600 match entries.
2. Current Score: Cumulative sum of runs grouped by match_id.
3. Balls Left: From ball parse overs and balls; compute balls_left = 120 - (overs_completed*6 + balls_in_over).
4. Wickets Left: Cumulative count of dismissals; then wickets_left = 10 - wickets_fallen.
5. Current Run Rate: current_run_rate = (current_score * 6) / balls_bowled.
6. Last 5 Overs Runs: Rolling sum over last 30 balls per match.
7. Final Runs: Total runs per match as the target variable.

Final features:

[batting_team, bowling_team, city,
 current_score, balls_left, wickets_left,
 current_run_rate, last_five]

Target: final_score (total runs).

---

Model Training

• Train/Test Split: 80% train, 20% test (random_state=42).
• Pipeline:
  1. ColumnTransformer:
     • OneHotEncode: batting_team, bowling_team, city
     • Passthrough numeric features
  2. StandardScaler for numeric columns
  3. XGBoostRegressor with:
     • n_estimators=1000
     • learning_rate=0.2
     • max_depth=12
     • random_state=1
• Evaluation:
  • R² Score: ~0.98
  • Mean Absolute Error: ~1.7 runs
• Artifact: Serialized pipeline saved to pipeline.pkl via pickle.

---

Installation

1. Clone the repository:

   git clone https://github.com/namaniisc/ScoreCast.git
   cd ScoreCast

2. Create a virtual environment & activate:

   python3 -m venv venv
   source venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

4. Ensure dataset and model:

   • T20_iore_info.csv in project root
   • pipeline.pkl in project root

---

Usage

1. Run the Flask app:

   python app.py

2. Access in browser: Navigate to http://localhost:5000.

3. Make Predictions:

   • Select teams and city.
   • Enter current match stats.
   • Click Predict Score to see the forecast.

---

Project Structure

cricket-score-predictor/
├── app.py               # Flask application
├── pipeline.pkl         # Serialized ML pipeline
├── T20_iore_info.csv    # Raw dataset
├── models/              # (Optional) model artifacts
├── notebooks/           # Jupyter notebooks
│   └── train_model.ipynb
├── requirements.txt     # Python dependencies
├── static/
│   └── images/          # Team flag images
└── templates/
    └── index.html       # Frontend template

---

Future Improvements

• Integrate real-time data APIs (e.g., Cricbuzz) for live inputs.
• Experiment with advanced architectures (LSTM sequences, ensemble methods).
• Add confidence intervals to predictions.
• Dockerize for containerized deployment.

---

Contributing

Contributions are welcome! Please:

1. Fork the repo
2. Create a feature branch (git checkout -b feature/YourFeature)
3. Commit your changes (git commit -m 'Add some feature')
4. Push to branch (git push origin feature/YourFeature)
5. Open a Pull Request