chadwickpy

Retrosheet play-by-play files to data tables. Pure Python, one pip install.

Retrosheet publishes every play of every MLB game as text files. The Chadwick tools turn those files into tables. chadwickpy is those same six tools, rewritten in pure Python: no compiler, no C library, no dependencies. Output is byte-identical to Chadwick on every Retrosheet season that has game data, 1897 to 2025 (for inputs where the C behaves in a defined way).

pip install chadwickpy            # or: uvx --from chadwickpy cwevent -h
curl -O https://www.retrosheet.org/events/2010seve.zip
unzip -q 2010seve.zip -d retro2010 && cd retro2010
cwevent -y 2010 -n -f 0,2,3,4,10,14,29,34 2010NYA.EVA > yankees_events.csv

"GAME_ID","INN_CT","BAT_HOME_ID","OUTS_CT","BAT_ID","PIT_ID","EVENT_TX","EVENT_CD"
"NYA201004130",1,0,0,"aybae001","petta001","S9/F9S-",20
"NYA201004130",1,0,0,"abreb001","petta001","K",3
"NYA201004130",1,0,1,"huntt001","petta001","8/F8LXD",2

The six tools

Same command names and options as Chadwick. Also usable from Python (from chadwickpy.tools.events import event_rows).

Good to know

• Speed: on one core chadwickpy is roughly 10 to 40 times slower than the C tools. With several files (a whole season) it uses several cores by itself: a full 2,430-game season takes about 2 seconds on a multi-core machine, the same as the C tools on one core. The output is identical at every core count (checked from 1 to 40). See the cores guide.
• Choose the number of workers with -j <workers> (-j 1 is sequential) or the CHADWICK_JOBS environment variable. By default it follows the CPUs the process may use, including a CPU limit set by Docker, Kubernetes or systemd.
• It reads Retrosheet's files; it does not include or download them.
• Pure Python, zero dependencies. Python 3.11 or newer; CI runs Python 3.11 to 3.13 on Linux, and it was also checked on 3.14. It should work on macOS and Windows too (on Windows it expands wildcards such as 2010*.EV* itself, as the C tools do), but those are not part of the automated tests.

Documentation

https://cbwinslow.github.io/chadwickpy/: getting started, a guide to each tool, Python usage, recipes, the full option and field reference, how it was verified, and an FAQ. Machine-readable: llms.txt.

Verification

Output is compared byte for byte (standard output, error messages and exit status) with the real Chadwick programs, built from the 0.11.0 release (tag v0.11.0):

• every season with game data, 1897 to 2025, all six tools and 16 option sets (2,064 comparisons);
• every output field of every tool on its own and in combination, both output formats, and every game-selection option;
• tests aimed at every remaining branch of the code (99% of the lines and branches were reached when measured on 0.3.0; the 0.4.0 suite of 5,163 tests passes against the real tools), and the rest are listed with the reason they cannot be;
• a weekly check against the newest Chadwick, and a test that fails when any C function changes.

Details and how to run them yourself: How it was verified. CI builds Chadwick and runs the whole suite on Python 3.11-3.13, failing on any skipped parity test.

Licence and credit

GPL-3.0-or-later (LICENSE). chadwickpy is a derivative work of Chadwick, Copyright (C) 2002-2026 Dr T L Turocy and the Chadwick Baseball Bureau, GPL-2.0-or-later (COPYING-chadwick, NOTICE); each ported module keeps that notice. It is independent: Chadwick and Retrosheet do not endorse or sponsor it, and it ships no Retrosheet data.

Retrosheet's data-use notice asks that anyone using its data say so. If you publish anything built on it, include:

The information used here was obtained free of charge from and is copyrighted by Retrosheet. Interested parties may contact Retrosheet at "www.retrosheet.org".

Contributing, questions, security

CONTRIBUTING.md · Discussions · Issues · SECURITY.md. The rule of the project: the port stays a translation of Chadwick's C; rules come from the C, never from tuning output.