• python
• pip

You may want to use a virtual environment.

• pip install -r requirements.txt

• main.py: execute this file to generate the decklist visualization
• savedecks.py: execute this file to crawl and save the current decklists of the cEDH decklist database
• cardstats.py: execute this file after savedecks.py to generate and save the card stats
• correlation.py: execute this file to generate the card visualization
• crawler.py: slightly modified copy of the cEDH Decklist Database crawler from https://anonymous.4open.science/r/cEDH-DDB-Crawler-D2E6
• cards.json: card attributes like type and mana cost from Scryfall, generate using mtgslides and copy here
• tree.pdf: generated file containing the visualization

These files are generated by executing scripts in the py folder above.

• links.json: Decklist links of the cEDH decklist database as a JSON array of decklist links, mostly from moxfield. Generated by py/savedecks.py.
• decks.json and decks.js: Decklists of the cEDH decklist database, Generated by py/savedecks.py.
• cards.json and cards.js: Card status of the cEDH decklist database. Generated by py/cardstats.py after py/savedecks.py

savedecks.py and cardstats.py are run daily in a GitHub workflow and deployed to data folder of the gh-pages branch.

If you want to generate the card visualization with correlation.py but don't want to install Python or the required libraries, you can use Docker as follows:

docker build . -t mtgcompare
docker run mtgcompare

• https://cedh-decklist-database.com
• https://www.moxfield.com The moxfield API is inofficial and may break at any time.

This is how I generate the monthly cEDH staples posts at https://reddit.com/r/cedh, for example see July 2022.

python py/savedecks.py
python py/cardstats.py

Or download https://github.com/KonradHoeffner/cedh/raw/gh-pages/data/cards.json to dist/data. Download the comparison file at that point in time from GitHub and save it to dist/data/cards.old.json.

python py/diff.py > diff.txt

Per request, only use the top two sections, top increased and top decreased.

1. if links.json does not exist:
   • fetch decklist links from the Competitive EDH Decklist Database using a crawler based on https://anonymous.4open.science/r/cEDH-DDB-Crawler-D2E6
   • save the result in links.json
2. if decks.json does not exist then for each moxfield decklist link:
   • replace https://www.moxfield.com/ with https://www.gerbrand.dev/
   • fetch the resulting URL
   • parse the content into a dictionary with decklist names as keys and lists of card names as values
   • save the result in decks.json
3. remove very common lands and mana rocks from the decks
4. transform the decks into binary vectors using a DictVectorizer
5. perform agglomerative clustering
6. create the linkage matrix
7. visualize the linkage matrix as a tree of labelled nodes

• use different distance functions
• use commander identity and card color relative scores in visualization
• add graph not based on clustering but on distance cutoff