This repository is a comprehensive fantasy football analytics platform built in R. It ingests player projections from over a dozen sources (CBS, ESPN, Yahoo, FantasyPros, etc.) via web scraping, transforms them into consensus estimates using robust statistics (Hodges-Lehmann estimator, weighted averages), and produces actionable insights such as player risk scores, value over replacement (VOR), optimal roster selections via linear programming, and draft-day bid-up-to recommendations. The pipeline also includes historical analysis modules that evaluate projection accuracy against actual Yahoo fantasy points, model expected values, and identify undervalued players ('gold mining'). The codebase is organized as a collection of R scripts with a clear separation of concerns: configuration (league settings, source definitions), reusable functions (data melting, value redistribution, scraping helpers), source-specific projection scrapers, calculation modules (risk, VOR, simulation, optimization), and post-processing for draft day and weekly roundups. Data is stored in both CSV and RData formats, with historical archives spanning 2008-2015. While the repository demonstrates solid data engineering patterns—externalized configuration, modular functions, robust statistical methods, and optimization—it suffers from significant maintenance and reproducibility issues. The scraping code is brittle, relying on hardcoded HTML structures that are likely broken. Dependencies are not pinned, and many packages (e.g., Rglpk, reshape, psych) may be unavailable or incompatible with modern R versions. The data is largely from 2013-2015, making it obsolete for current NFL analysis. There are no tests, and manual steps are required for some data acquisition. As a result, the repository is best treated as a historical reference and a source of reusable patterns rather than a production-ready system.

The LLM identified 5 files worth highlighting for reuse: