MTG PROJECT

For my Magic: The Gathering forecasting project, I developed machine learning models to predict card prices, popularity, and power levels. I began by creating a custom data collection platform at MTGData.com, where users rated cards on subjective qualities like playability and versatility. After gathering responses for 250 cards, I supplemented this dataset with comprehensive card information from the Scryfall API, including text, rarity, and image data. This dual-source approach provided both quantitative metrics and qualitative assessments to train my models.

My methodology focused on creating a quantitative framework for evaluating Magic cards through systematic analysis. I processed the combined datasets into a machine-readable format, then trained models to identify patterns that determine a card's market value and playability. Additionally, I developed a framework using statistical analysis to assess whether cards were over or underpowered, creating bell curves and distribution analyses to establish benchmarks for card evaluation. The final models were designed to provide actionable insights for both players and card designers, with plans to implement them as Discord bots and APIs for community use.