MLB Game Predictions — SageMaker AI Demo

Today's Game Predictions

Predictions are generated each day at noon ET using a retrained XGBoost model. Features include team win rates, recent form, home/away splits, runs scored/allowed, and starting pitcher ERA, WHIP, and K/9.

Loading predictions…

Ask the Deployment

Claude via Bedrock

Ask any question about how this demo works — the SageMaker model, the AWS infrastructure, the features, the pipeline, anything.

Hi! I'm an AI assistant that knows everything about how this MLB prediction system is built. Ask me about SageMaker, XGBoost, the AWS infrastructure, the feature engineering, the daily retraining pipeline — anything!

Architecture Overview

🗄️

Data Source

Official MLB Stats API (free, no key required). Fetches 3 seasons of game results plus starting pitcher game logs.

🔧

Feature Engineering

20 rolling features built with zero data leakage: team win rates, form, splits, runs, and pitcher ERA/WHIP/K9.

🧠

SageMaker XGBoost

Built-in managed container. Training job, Model, Endpoint Config, and Endpoint all visible in the AWS Console.

☁️

Web Infrastructure

S3 + CloudFront + ACM for HTTPS. Lambda + API Gateway for dynamic content. Route 53 for DNS.

🔄

Daily Retraining

EventBridge fires at noon ET daily. Lambda fetches yesterday's results, retrains, updates the endpoint.

🤖

Chatbot

Amazon Bedrock (Claude Haiku 4.5) reads projectknowledge.md to answer student questions about the deployment.