🏏 Cricket Match Winner Prediction (IPL & T20)

This project predicts the winner between two cricket teams for IPL and T20 International matches using machine learning. It combines real-world cricket data, advanced engineered features, and classification models to generate accurate predictions and interactive dashboard outputs.

---

🚀 Features

• 🔮 Predict winners using pre-match team and venue context
• ⚙️ Trained XGBoost models using player and team statistics
• 📊 Streamlit dashboards for IPL, T20, and a unified view
• 📥 Feature-rich pipelines for IPL and T20 data with independent training

---

📊 Engineered Features

T20 Features (Advanced Player-Level)

• t1_form, t2_form: Rolling avg of last 5 matches (batting/wicket form)
• t1_win_rate_overall, t2_win_rate_overall: Historical performance
• batsman_vs_bowler: Head-to-head stats
• venue_familiarity: Runs/wickets at a given venue
• h2h_win_rate: Win % of team1 over team2
• toss_factor: Fixed to 0.5 due to missing toss data

IPL Features (Team-Level Stats)

• t1_h2h_win_rate: Historical win rate of team1 vs team2
• t1_toss_advantage: How often team1 wins when it wins the toss
• t1_venue_win_rate: team1 win rate at a venue
• Encoded: team1, team2, venue

---

🛠️ Setup Instructions

1. Move to project folder:

cd cap5771sp25-project

2. Create virtual environment and activate it:

python3 -m venv .venv
source .venv/bin/activate  # For macOS/Linux
.venv\Scripts\activate   # For Windows

3. Install dependencies:

pip install -r requirements.txt

4. Run T20 dashboard:

streamlit run dashboard_it20.py

5. Run IPL dashboard:

cd ipl
source .venv/bin/activate
pip install -r requirements.txt
python3 dashboard.py

6. Run Combined Dashboard:

python3 dashboard_combined.py

---

📉 Evaluation Metrics (Milestone 3)

• Accuracy: 87%
• F1 Score: 87%
• ROC AUC: 96%

---

📉 Evaluation Metrics (Milestone 2)

📈 IPL Results

Model	Accuracy	Precision	Recall	F1 Score
Logistic Regression	0.761	0.764	0.750	0.757
Random Forest	0.642	0.636	0.648	0.642
XGBoost	0.619	0.619	0.602	0.610

🌍 T20 Results

Model	Accuracy	Precision	Recall	F1 Score
Logistic Regression	0.653	0.594	0.679	0.634
XGBoost	0.648	0.595	0.635	0.615
Random Forest	0.643	0.597	0.586	0.591

---

💡 What is Working

• ✅ Separate pipelines for IPL and T20 datasets
• ✅ Custom preprocess.py for rolling features, venue stats, and matchup insights
• ✅ .pkl model saving and reloading for consistent reuse
• ✅ Streamlit dashboards: dashboard_it20.py, dashboard.py, and dashboard_combined.py
• ✅ Same metrics used in training and test set for evaluation
• ✅ All major features engineered and integrated

---

📦 Tools & Tech Stack

Category	Tools Used
Programming	Python
Modeling	XGBoost, Scikit-learn
UI & Visualization	Streamlit, Matplotlib, Seaborn
Data	Pandas, NumPy
Reporting	fpdf (planned: auto PDF generation)

---

📄 Milestones

✅ Milestone 1:

• Original project (Airline Delay Prediction)

✅ Milestone 2:

• Switched to cricket prediction
• Team-level stats, encoded variables
• Used Logistic Regression, Random Forest, XGBoost

✅ Milestone 3:

• Player-level stats, rolling form, and venue familiarity
• Optimized XGBoost, 87% accuracy
• Streamlit dashboards per format + combined view
• PDF reporting planned

---

👥 Team

• GUTTAPATI JAYASURYA REDDY (T20 Pipeline & Dashboard)
• LOKESH MAKINENI (IPL Pipeline & Integration)

---

🎥 Video Links

4-minute PPT presentation: https://youtu.be/s9mcpF-C7W4

4-minute demo: https://youtu.be/bA_75s23IV0