Chirag Suri

Aspiring Data Analyst - Passionate about insights, dashboards, and the game of cricket.

Back to Projects

In this project, I dove deep into understanding what separates an ordinary striker from a great one. Using Python, I explored a dataset of 500 strikers containing personal attributes and on-field performance stats — all with the goal of answering one simple question:

👉 What makes a football striker exceptional?

🛠️ Tools & Technologies Used

  • Python (pandas, seaborn, matplotlib, scipy, sklearn, statsmodels) – For EDA, visualization, statistical testing, clustering, and machine learning.
  • Jupyter Notebook – For code execution and documentation.
  • Generative AI (ChatGPT) – Helped refine analysis steps, validate statistical logic, helped with ideation and structure the project cleanly.

🔗 Links

GitHub LinkedIn Portfolio

📁 Dataset

I worked with a dataset of 500 football strikers, covering both demographic and performance-based variables.

📌 Note: I couldn’t locate the original source, so it’s being treated as synthetic/simulated data for educational purposes.

🎯 Project Objectives

This project was focused on segmenting and classifying strikers based on their:

  • ⚽ On-field performance (Goals, Assists, Dribbling, etc.)
  • 🧠 Attributes (Footedness, Consistency, Game IQ, Conduct)
  • 📊 Team Impact & Match Influence

The broader goal was to develop a data-backed system for:

  • Identifying top-tier vs average strikers.
  • Helping scouts/coaches with player selection.
  • Answering questions even analysts may overlook.

🧠 Key Questions Solved

  • What’s the maximum number of goals scored by a striker?
  • What percentage of strikers are right-footed?
  • Which nationality scores the most goals on average?
  • What’s the average conversion rate of left-footed players?
  • Do hold-up play skills correlate with consistency?
  • Are consistency scores normally distributed?
  • Is there a statistical difference in performance between nationalities?
  • Can we predict striker types using logistic regression?

🔄 Project Workflow

🧼 1. Data Cleaning & Preparation

  • Handled null values using SimpleImputer:
    • Median for numeric
    • Most frequent for categorical
  • Typecasting of key performance metrics (e.g., Goals, Assists).
  • Used LabelEncoder for footedness and marital status.
  • Created dummy variables for nationality.

📊 2. Exploratory Data Analysis

  • Descriptive stats for all key metrics.
  • Pie chart for footedness distribution.
  • Countplot of footedness by nationality.

📈 3. Statistical Analysis

  • Used groupby + mean to analyze national scoring rates.
  • Performed Shapiro-Wilk for normality check.
  • Levene’s Test to validate homogeneity before ANOVA.
  • Correlation (Pearson) and regression to understand how “Hold-up Play” influences consistency.

🧪 4. Feature Engineering

  • Created a Total Contribution Score from key fields:
    • Goals, Assists, Shots on Target, Dribbles, etc.
  • Used this score for clustering and ML input.

🧠 5. K-Means Clustering

  • Identified 2 clusters using elbow method.
  • Tagged strikers as:
    • Best Strikers
    • Regular Strikers

🤖 6. Machine Learning (Logistic Regression)

  • Trained model to classify striker type.
  • Used StandardScaler for normalization.
  • Achieved solid accuracy and visualized predictions via confusion matrix.
View Notebook

💡 Key Insights

  • 🥇 The highest individual goal tally: 36 goals
  • 🦵 Right-footed strikers dominate the dataset (about 73%)
  • 🌍 Brazilian strikers had the highest average goal count
  • 🧠 Hold-up Play shows positive correlation (0.55) with consistency
  • 📊 Consistency scores were not normally distributed, but heteroscedasticity was not an issue
  • 🧪 Regression model showed hold-up play significantly predicts consistency
  • 🧮 Best strikers had an average contribution score of ~212
  • ✅ Logistic Regression model achieved X% accuracy

📚 Things I Learned

  • ✅ End-to-end structuring of ML-based projects
  • 📊 Choosing the right statistical test depending on data assumptions
  • 🧹 Proper preprocessing: imputation, encoding, scaling
  • 💬 Explaining results with storytelling, not just numbers
  • 🤝 Using generative AI to validate and speed up analysis

🚀 How to Explore

If you’re looking to test the code:

  1. 📥 Clone or download the repo
  2. 🐍 Open Football-Striker.ipynb in Jupyter
  3. 🧠 Run through sections, tweak assumptions, explore clusters and model results

✅ What’s Next?

  • Building a Power BI Dashboard to turn this analysis into a visual scouting tool

🙏 THANK YOU! 🙌

Back to Projects