Boba | Discover 2025 MiLB Prospects

Boba

Inspiration

When the idea of a popularity leaderboard was suggested, that convinced us to settle on Challenge 5 for a prospects leaderboard instead. There were similar platforms like FanGraphs, but none attempted using machine learning for prospect predictions.

What it does

This is a prospects leaderboard that uses AI-predicted WAR scores to rank MiLB prospects.

How we built it

The web-app was built with Next.js using TypeScript and shadcn/ui for the frontend. The data was retrieved using the MLB-StatsAPI Python wrapper library and stored on Firebase. The model was created with Vertex AutoML (98% R2), which we used due to its superior accuracy over H2O AutoML Stacked Ensemble (97% R2).

Challenges we ran into

Feature selection for the machine learning pipeline took time and deliberation before running the actual training sequence.
Retrieving sabermetrics data with the StatsCast API required spending time reading documentation on the required query parameters to retrieve the proper data.
Data cleaning for remove non-pitchers that were included in the pitcher queries.
Originally, Chakra UI was selected, but shadcn/ui offered a smoother development experience for using CLI inputs or manual copy/pasting components.
Parsing the .json schemas to template into the table required careful attention to resolve errors.

Accomplishments that we're proud of

Achieving 98% MLB WAR accuracy based on the feature selection we settled on.
Integrating a reusable component library like shadcn/ui helped speed up development.
Working application that displays all data with player and team images.

What we learned

Storing images as compressed base64 strings in Firebase to avoid running costs of standard image storage.
Learning about WAR as the gold-standard of representing overall player value.

What's next for Boba | Your AI-Powered MLB Superfan

Potential to integrate other positions (hitters, fielders) and other leagues (e.g., international league).
Potential to select other languages for the web app.