nba-clv-dashboard

FastAPI + Chart.js evaluation dashboard for calibration, rolling accuracy, and CLV-style reporting.

Why This Repo Matters

Most public ML repos stop at a headline metric. This one is about how to communicate whether a model is actually trustworthy:

• calibration instead of just raw accuracy
• rolling performance instead of one static score
• CLV-style reporting to talk about market-aware outcomes
• clean API + frontend surface for presenting evaluation to non-ML stakeholders

One-pager for recruiters: Sports ML evaluation case study

What It Does

FastAPI serves /api/metrics and the frontend renders:

• calibration scatter
• rolling accuracy chart
• KPI/summary block for CLV-style metrics
• a lightweight, portable demo for model evaluation storytelling

Quick Start

pip install -e .
uvicorn nba_clv_dashboard.app:app --reload --port 8765

Open http://127.0.0.1:8765.

Swap Your Metrics

Replace demo_metrics.demo_payload() with a loader from your parquet or evaluation pipeline. Keep the same JSON shape or adjust static/index.html.

Non-goals

• Not a sportsbook integration
• Not a multi-tenant analytics product
• Not a prediction engine by itself

CI

pytest hits /api/metrics with httpx.

Related Repos

• nba-ratings: reusable Elo / logistic / Kelly primitives
• sports-betting-ml: modeling demo and value-bet pipeline
• backtest-report-gen: static report generation from evaluation JSON

License

MIT