MLB Aapp

Inspiration

As a machine learning enthusiast and sports analytics fan, I’ve always been fascinated by the way data transforms decision-making in sports. But my real motivation for this project came from two key moments—one from my personal journey and another from the world of baseball itself.

I remember watching a rookie player during an MLB game—his bat speed, exit velocity, and launch angle were impressive, but he wasn’t considered a top-tier prospect. He hit several near-home runs, yet scouts weren’t convinced. I thought:

💡 "What if data could tell a different story? What if we could predict a player's true potential—not just based on hype, but on real, deep insights from historical comparisons?"

That question sparked an idea: Could machine learning be used to evaluate a player’s future success based on advanced metrics?

What it does

This project is a deep learning-powered platform that serves two main purposes: Uses advanced Statcast metrics (Exit Velocity, Launch Angle, Hit Distance, etc.) to evaluate a player's potential. Employs deep learning (LSTM model) to analyze historical player data and predict future career success. Helps coaches, scouts, and analysts identify underrated talents and make data-driven decisions.

How we built it

Challenges we ran into

Accomplishments that we're proud of

This project combines deep learning, real-time analytics, and API-based data collection to predict MLB prospects' future success and provide live baseball strategy insights. Data Source: We fetched Statcast data from MLB's public datasets, which include:

Exit Velocity, Launch Angle, Hit Distance (key batting metrics) Pitch Speed, Pitch Type (for analyzing pitching strategy)

We trained a Long Short-Term Memory (LSTM) model, a type of recurrent neural network (RNN) well-suited for analyzing time-series baseball performance data.

LSTM layers for capturing sequential patterns in player performance. Dense layers with ReLU activation to enhance feature extraction. Sigmoid output for predicting probabilities (e.g., likelihood of a home run). Binary Crossentropy loss for classification tasks.

Trained on past MLB player performance data, using historical comparisons to predict future success. Optimized with Adam optimizer for faster convergence. Validation with test data to ensure generalization.

Built With

• deep
• fastapi
• learning
• python
• tensorflow