Boba | Discover 2025 MiLB Prospects with 98% Accuracy

Boba is a Next.js application that helps users discover Minor League Baseball (MiLB) prospects using a machine learning model with 98% predictive accuracy (R²). It provides a user-friendly interface to explore the most recent 2025 MiLB prospect pool, offering detailed prospect profiles and rankings. The app leverages Next.js, Firebase, and Tailwind CSS for a modern and efficient experience.

Contributors

Developer	Affiliation	Contact
Ebod Shojaei	BSc. University of British Columbia	ebod.shojaei@alumni.ubc.ca
Rebecca Jeon	BSc. University of Victoria	beccajeon12@gmail.com

Features

• View a comprehensive list of all 2025 MiLB prospects.
• Explore individual MiLB prospect profiles.
• Sort MiLB prospects by ranking.

Technologies

• Frontend: Next.js, Tailwind CSS
• Backend: Firebase

Routes

• /: Home page showcasing a list of 2025 MiLB prospects.
• /about: About page with details on the MLB testing data and the "bWAR" metric.
• /contact: Contact page for inquiries, feedback, or bug reports.

Data and Methodology

Data is sourced from the MLB Stats API. Our proprietary machine learning model (Vertex AI) was trained on over 6,000 players and rigorously tested on over 1,500 players from the 2015 to 2024 seasons. This model predicts WAR (Wins Above Replacement), which we name "bWAR" (Boba Wins Above Replacement), for over 600 available 2025 prospects.

Our "WAR Machine" achieves 98% accuracy (R²) in predicting MiLB prospect WAR based on testing. Error rates were calculated using a modified Symmetric Mean Absolute Percentage Error (sMAPE) for two values, indicating both magnitude and direction within a range of -100% to 100% (0% sMAPE is perfect accuracy). Change in bWAR was calculated for players with available 2023 data to indicate growth or decline.

The model training process included:

1. Data Loading
2. Data Splitting (Train/Test)
3. Data Cleaning
4. Missing Value Imputation
5. Feature Scaling
6. Model Training (using Vertex AI AutoML)
7. Model Evaluation
8. Prediction on 2024 MiLB data
9. Model Saving
10. Prediction Saving

A subset of the following pitching statistics from the MLB Stats API were used as features for model training:

Field	Description
stat.gamesPlayed	Number of games pitched.
stat.gamesStarted	Number of games started.
stat.gamesFinished	Number of games finished.
stat.completeGames	Number of complete games.
stat.shutouts	Number of shutouts.
stat.wins	Number of wins.
stat.losses	Number of losses.
stat.saveOpportunities	Number of save opportunities.
stat.saves	Number of saves.
stat.blownSaves	Number of blown saves.
stat.holds	Number of holds.
stat.inningsPitched	Innings pitched (can be a formatted string).
stat.runs	Runs allowed.
stat.earnedRuns	Earned runs allowed.
stat.battersFaced	Number of batters faced.
stat.atBats	At-bats against the pitcher.
stat.hits	Hits allowed.
stat.doubles	Doubles allowed.
stat.triples	Triples allowed.
stat.homeRuns	Home runs allowed.
stat.baseOnBalls	Walks issued.
stat.intentionalWalks	Intentional walks issued.
stat.strikeOuts	Strikeouts.
stat.hitByPitch	Batters hit by pitch.
stat.balks	Balks committed.
stat.wildPitches	Wild pitches.
stat.groundOuts	Groundouts induced.
stat.airOuts	Flyouts induced.
stat.stolenBases	Stolen bases allowed.
stat.caughtStealing	Runners caught stealing.
stat.sacBunts	Sacrifice bunts allowed.
stat.sacFlies	Sacrifice flies allowed.
stat.catchersInterference	Catcher's interference while pitching.
stat.pickoffs	Pickoffs.
stat.inheritedRunners	Inherited runners.
stat.inheritedRunnersScored	Inherited runners scored.
stat.numberOfPitches	Pitches thrown.
stat.strikes	Strikes thrown.

Acknowledgements

• MLB Stats API

---

Disclaimer:

• This project was created for submission to the 2025 Google X MLB Hackathon.
• We are not affiliated with the MLB or any of its partners.
• AI predictions are not guaranteed to be accurate.
• We are not responsible for any errors or inaccuracies in the data.
• The use of AI was for debugging purposes only.