Scrabble Player Rating — Machine Learning Ranking System

• Tools & Technologies: Python, Pandas, Scikit-Learn, XGBoost, Random Forest, Linear Regression, Data Visualization
• Role: Data Scientist
• Github URL: Project Link

Description:

Analyzed over 73,000 competitively logged Scrabble games from Woogles.io to design a statistically robust ML-based player-rating algorithm. Performed extensive data engineering including text-derived feature extraction, outcome normalization, win-probability modeling, and turn-by-turn EDA. Evaluated three algorithms under cross-validation to measure predictive stability, calibration curves, and performance variance across skill tiers. Developed an interpretable model schema capable of quantifying skill progression over time.

Outcome

XGBoost delivered the strongest predictive performance with significantly lower RMSE than baseline models. Results validated the feasibility of replacing traditional Elo-based systems with dynamic Machine Learning (ML) driven rating strategies.