here is an image of the product and the link is in the discription!

🏏 Cricket Win Predictor

A practical IPL/T20 prediction starter project with:

• Data foundation (match, player, venue, team-form, and ball-by-ball CSVs)
• Feature-ready training frame for chase prediction
• Streamlit web app to explore datasets and predict outcomes

---

✨ What this repository provides

• Sample IPL datasets across multiple granularities (match-level, player-level, and delivery-level)
• Starter chase-state training data (final_df.csv)
• Streamlit app with:
  • Predict Winner workflow
  • Data Explorer workflow for all CSVs in data/
  • Heuristic fallback prediction if models/pipe.pkl is not available

---

🗂️ Project structure

CRICKET_PREDICTOR/
│
├── app/
│   └── app.py
├── data/
│   ├── raw_data.csv
│   ├── final_df.csv
│   ├── player_stats.csv
│   ├── matches.csv
│   ├── deliveries.csv
│   ├── players.csv
│   ├── batting_stats.csv
│   ├── bowling_stats.csv
│   ├── venues.csv
│   ├── head_to_head.csv
│   ├── team_form.csv
│   ├── team_strength_metrics.csv
│   └── match_conditions.csv
├── notebooks/
│   └── data_cleaning.ipynb
├── models/
│   └── pipe.pkl              # optional (trained model)
├── requirements.txt
└── README.md

---

📊 Dataset guide

1) Match-level data

• data/matches.csv
• Key fields include: match_id, season, city, venue, team1, team2, toss_winner, winner, team1_score, team2_score

2) Ball-by-ball data

• data/deliveries.csv
• Key fields include: match_id, inning, over, ball, batsman, bowler, runs_scored, extras, wicket

3) Player profile data

• data/players.csv, data/player_stats.csv
• Includes role, batting/bowling style, team, experience, and lifetime indicators

4) Player performance data

• data/batting_stats.csv
• data/bowling_stats.csv
• Includes strike-rate/economy and phase-relevant metrics

5) Team & venue context

• data/venues.csv
• data/head_to_head.csv
• data/team_form.csv
• data/team_strength_metrics.csv
• data/match_conditions.csv

These are useful for engineered features like:

• toss impact
• venue advantage
• recent form difference
• head-to-head ratio
• batting vs bowling strength gap

---

🧠 Modeling blueprint (recommended)

1. Start with raw_data.csv / deliveries.csv.
2. Build ball-by-ball chase-state features.
3. Use final_df.csv as a baseline schema.
4. Join context data (venue/team form/head-to-head/player strength).
5. Split train/test by match or season (avoid leakage).
6. Train baseline Logistic Regression and tree models (Random Forest / XGBoost).
7. Save the best pipeline to models/pipe.pkl.

---

🚀 Run the app

python scripts/train_model.py

App capabilities

• Predict Winner tab

  • Enter batting team, bowling team, city, target, score, overs, wickets
  • See win probability for both teams
  • View derived match-state features + engineered context features

• Data Explorer tab

  • Browse all data/*.csv
  • Inspect rows/columns in-app
  • Download selected CSV directly

If models/pipe.pkl does not exist, the app still predicts using a built-in heuristic fallback.

App capabilities

• Predict Winner tab
  • Enter batting team, bowling team, city, target, score, overs, wickets
  • See win probability for both teams
  • View derived match-state features + engineered context features

✅ Next steps

• Replace sample CSVs with full historical IPL datasets.
• Add training scripts (scripts/feature_engineering.py, scripts/train_model.py).
• Evaluate with season-wise validation.
• Persist the best model to models/pipe.pkl.

---

📦 Core dependencies

• streamlit
• pandas
• numpy
• scikit-learn
• xgboost
• jupyter

---

📝 License

This project uses the repository's LICENSE file.