AI NFL GM

A Carleton AI Society project. We want to price every NFL player and draft pick in one number, surplus value: what a player produces minus what he costs against the salary cap. From that we build a trade calculator and a draft board for carletonai.com. All the data is free.

We have not started building yet. For now the job is to get the data on your laptop and look around.

Get the data

Install uv (it brings its own Python), then:

git clone https://github.com/carletonai/ai-nfl-gm.git
cd ai-nfl-gm
uv run get_data.py   # about 20 MB into data/, takes a minute or two
uv run jupyter lab   # start a notebook

In the notebook:

import pandas as pd

contracts = pd.read_parquet("data/contracts.parquet")
contracts[["player", "position", "year_signed", "apy_cap_pct"]].head()

What's in data/

File	What it is
contracts	Every contract Over The Cap tracks. apy_cap_pct is the yearly value as a share of the cap
players	One row per player, with the IDs that join the other tables (gsis_id, pfr_id, ...)
draft_picks	Every pick since 1980: team, slot, college, career value (w_av)
combine	Combine results since 2000: 40 time, bench, vertical, ...
trades	Trades since 2002
player_stats	Regular-season stats per player per season, 2011 to now

Two things that trip people up: each contract's year-by-year cap hits sit inside its season_history column, and car_av in draft_picks is empty, so use w_av.

Questions

Ask in #ai-nfl-gm on the CAIS Discord, or ask Jeremy.

---

Code is MIT. The data comes from nflverse, which packages Over The Cap and Pro Football Reference data. Credit them, and keep the downloaded files out of git (data/ is ignored already).