TWIDDDS

Inspiration

This project was inspired by my goal to strengthen my skills in data science and machine learning. I wanted a practical, hands-on way to apply concepts like data cleaning, visualization, and predictive modeling. Although I haven’t played the game behind the data, it served as a rich context for experimentation and learning.

What it does

This project takes in champion statistics from two consecutive game patches and analyzes how key metrics—like win rate, pick rate, ban percentage, and KDA—change over time. Using this information, it predicts whether each champion is likely to be Buffed, Nerfed, or remain Neutral in the upcoming patch. It also generates visualizations to explore relationships between variables and helps make sense of game balancing trends.

How we built it

• Merged and organized patch data, handled missing values, and ensured consistent formatting across features.
• Plotted scatterplots, density graphs, and linear regression visuals to understand trends and distributions.
• Trained the model on current patch features to classify changes as Buff, Nerf, or Neutral in the next patch.
• Evaluated the model and identified which stats were most influential in the predictions.

Challenges we ran into

• Cleaning and transforming real-world data while maintaining consistency across patches.
• Interpreting and resolving R errors related to data types.
• Ensuring correct alignment between feature columns from different patches.
• Making the model generalize well across different scenarios with limited data.

Accomplishments that we're proud of

• Successfully applied machine learning techniques on real-world-style data.
• Debugged tricky issues with R, including handling data types and model input formats.
• Gained hands-on experience in scaling features and building a complete analysis pipeline.
• Created clean, meaningful visualizations that reveal patterns in champion performance.
• Completed a full data science project independently, from data cleaning to modeling and interpretation.

What we learned

• Build and evaluate a Random Forest classification model
• Use R for data visualization
• Interpret model predictions and communicate insights

Built With

• r