NBA Analytics App

A real-time NBA analytics platform that leverages automated data ingestion and custom projection models to provide actionable insights for evaluating player prop bets.

Explanation:

The sports betting space is heavily driven by data, but most tools either rely on surface-level stats or require manual analysis. I set out to build a system that could ingest real-time National Basketball Association data and turn it into actionable insights for evaluating player prop bets.

Challenges and Key Features:

• Early on, I underestimated the complexity of sourcing reliable data.
• Attempted to scrape multiple websites for stats and odds
• Ran into inconsistent data formats and mismatched schemas
• Faced accuracy concerns due to unreliable or partial data
• Spent significant time cleaning data instead of building features

Implementation Details:

I shifted from scraping to a structured, API-driven architecture. Data Source Upgrade: Transitioned from web scraping to a paid API for consistent, structured data. Data Architecture: Designed a raw → analytics (staged) pipeline. Raw layer stores untouched API data for auditing; Analytics layer transforms data into query-ready structures. Automation: Built scheduled ingestion pipelines using AWS Lambda + EventBridge managed via Terraform CLI. Modeling: Developed baseline projection models using historical performance and calculated EV vs market odds.

What I Learned:

• Data quality > everything: Scraping introduced too much inconsistency — reliable APIs made everything downstream easier.
• Schema design should come first: Planning the data model upfront saves significant time.
• Separate raw and processed data: This allowed safe transformations without risking data integrity.
• Trust your data pipeline before your models: Doubts in data accuracy directly impact confidence in outputs.
• Real-world tooling matters: Gained hands-on experience with AWS Lambda, EventBridge, and Terraform through CLI-based workflows.

Related Blog Posts: