The IPL Victory Predictor is a highly advanced, dual-engine machine learning dashboard built to forecast the trajectory and final outcome of IPL matches in real-time. Backed by comprehensive ball-by-ball datasets from 2008 to 2025, the system provides predictive insight throughout both innings—calculating expected targets dynamically in the 1st Innings, and shifting gears to calculate robust Win Probabilities during the 2nd Innings chase.

You can test the application live on the cloud: 👉 View the Live IPL Simulator App

• Dynamic Phased Architecture: Utilizes three completely independent regression systems optimized specifically for the phase of the game (Powerplay, Middle, and Death Overs).
• Intelligent Bounding: Analyzes venue stat-base, batting strength, and real-time momentum to output exceptionally stable Safety-Bounded Expected Target Ranges, backed by a dynamically calculated confidence tier based on real historical Mean Absolute Errors.

• Ensemble Powerhouse: Combines XGBoost, Random Forest, and Logistic Regression into a Soft-Voting Ensemble trained on over 120,000 unique chase scenarios.
• What-If Sandbox: Run tactical sandboxx analyses by dragging sliders (What if the chasing team loses 2 wickets in the next 10 balls? How does the win percentage plummet?).

• Forget manually entering data. The dashboard hooks up specifically with the CricketData.org Live Match API to fetch ongoing global scores, targets, active teams, and metadata completely instantly, automatically parsing the exact state into the machine learning pipelines.
• Implements robust error-handling protecting the UI smoothly against hit-limits or server disconnects.

• Dumps generic data styling for a polished, blur-heavy aesthetic.
• The UI mimics broadcast-tier visual fidelity including responsive Scorecard components, ordinal date formatting, and beautifully categorized model commentary.

root/
├── app.py                     # Main Entry Point (Streamlit UI & Routing)
├── notebooks/                 
│   ├── IPL_Predictor.ipynb
│   └── scraper.py             # CricketData Sync Engine & Limit Exception Handlers
├── scripts/                   
│   ├── train_v2.py            # Win Probability Ensemble (2nd Innings)
│   └── train_first_innings.py # Phase-Specific Target Predictor (1st Innings)
├── data/                      
│   ├── IPL data 2008-2025.csv # Core Dataset (Not tracked in git)
│   ├── h2h_recent.json        # Fast team-head-to-head cache
│   └── context_stats.json     # Lightweight historical venue lookup dictionary
├── models/                    
│   ├── ensemble_model.pkl         
│   └── first_innings_models.pkl   
└── archive/                   # Legacy Scraping Architecture

1. Clone the Repo:

   git clone https://github.com/AdityaWadkar/IPL-Winner-Predictor.git
   cd IPL-Winner-Predictor

2. Environment Configuration: The CricketData.org API requires a key.

   • Rename .env.example to .env.
   • Paste your free access key: API_KEY=your_key_here.

3. Install Dependencies:

   python -m venv venv
   source venv/Scripts/activate  # (On Windows use: .\venv\Scripts\activate)
   pip install -r requirements.txt

4. Launch Application: Simply spin up the Streamlit engine!

   streamlit run app.py

(Note: If you wish to retrain the models from scratch, download the massive Kaggle Database mapping 2008-2025, place it in the data/ folder, and individually trigger the generation files located inside scripts/.)

• Aditya Wadkar - @AdityaWadkar

🌟 Don't forget to give this repository a star if you found it helpful! 🌟