Golden Scout

Player Comparison
Executive Report
Decision
Trade Simulator
Season Simulation

Inspiration

NBA contracts are worth hundreds of millions of dollars, yet most decisions are made on gut instinct and highlight reels. We wanted to build a tool that brings the same rigor a quant fund applies to stocks — to basketball players.

What it does

Golden Scout is an NBA contract decision engine. Enter any player and a salary ask — it runs position-adjusted z-scores, 10,000 Monte Carlo season simulations, and a custom durability model (GP/82 × age curve) to compute a health-adjusted fair value. The verdict: SIGN, NEGOTIATE, or AVOID, with every number explained. Features include a Trade Impact Simulator, head-to-head Player Comparison with AI verdict, and an AI Executive Report read aloud via ElevenLabs.

How we built it

React + Vite frontend, Python FastAPI backend, Pandas for stat aggregation from 2024-25 NBA data, Google Gemini 2.0 Flash for AI reports and compare verdicts, ElevenLabs for voice output, and a Monte Carlo simulation engine running 10,000 seasons per player.

Challenges we ran into

Getting accurate games played and age data required merging multiple CSVs and preserving values through aggregation pipelines. Fuzzy name matching for Unicode player names (Jokić, Dončić) needed careful handling. Calibrating the durability model so injury risk was meaningful but not overwhelming took significant iteration.

What we learned

Translating academic sports economics (Berri & Schmidt 2010) into a real product pipeline. How to chain multiple AI services (Gemini + ElevenLabs) with caching so the app stays fast. Building a Monte Carlo engine that's both statistically sound and explainable to non-technical users.