mattgilgo/fantasy_football ? reverse-engineered prompt

Build me a fantasy football projection project that trains on historical NFL player stats and spits out draft rankings for QB, RB, WR, and TE.

I want it to pull in the usual data sources like player stats, target projections, expert consensus rankings, and ADP, then compare a few different regression models and pick the best one for each position. Please include a way to retrain for a new season, generate updated rankings, and save the results as CSV files I can use before my draft. It would also be great if it flags players who look like true stars at their position and shows how the models compare against expert projections using MAE.

Use notebooks for the workflow, with clear steps for exploring the data, training models, and reviewing results after the season. If you need current library details or best practices, look them up online first.