How the Wikidata Card Game Generator Chooses Cards and Properties

Updated

The generator picks a topic's Wikidata items, counts which properties appear most often across them, then keeps the 32 items that fill in the most of those frequent properties. It works on any topic that exists as a Wikidata entity with subclasses — you supply the entity ID (the Q-number), and the tool does the rest. Cards can end up missing data or images when the underlying Wikidata items are incomplete, which is a data problem, not a bug.

The three-step selection process

The site describes its own logic in three stages:

  1. Fetch items by subclass. It pulls items from Wikidata that are instances of arbitrary subclasses of the topic you provided. So you don't get only direct instances — the net is cast across the topic's subclass tree.
  2. Find the most common properties. Among those fetched items, it counts which properties occur most often and treats those as the card stats.
  3. Select 32 cards. From the pool, it chooses the 32 items that have as many of those frequent properties filled in as possible.

The result is a Top Trumps–style deck where each card's stats are Wikidata properties, and the 32 cards are the best-documented items in that topic.

What you need to provide

You give the generator a single Wikidata entity ID — the Q-number for your topic. The site's instructions:

  • Think of a topic and type its singular form into the search box on Wikidata.
  • Find the entity ID near the top of the page; it starts with Q.
  • Paste that ID into the generator's box and click Generate.

For example, the ID of chemical element is Q11344.

Worked example

Say you want a deck about chemical elements:

  1. Search Wikidata for "chemical element" (singular).
  2. Note the entity ID: Q11344.
  3. Paste Q11344 into the generator and click Generate.

The generator then gathers items that are instances of subclasses of chemical element, finds the properties most commonly filled in across them, and returns the 32 best-populated items as cards.

Topics the site suggests trying

The generator lists these as starting points:

  • capital
  • metro system
  • country
  • astronomical body
  • island
  • hackerspace
  • computer model
  • telescope
  • card game

Each of these has enough subclass structure and populated properties in Wikidata to produce a full deck.

Why cards may be missing data or images

Because selection depends on how completely items are documented in Wikidata, gaps in the source data show up as gaps on the cards. The site's own suggestion is direct: if your card is missing data or images, add that information to Wikidata. Improving the underlying item improves the deck.

Practical implications

  • Topic choice matters. A topic with many well-documented subclasses yields richer cards than a sparse one.
  • Property frequency drives the stats. The stats on your cards aren't hand-picked; they're whatever properties the tool found most often across the fetched items.
  • 32 is fixed. The deck size is set, so the tool optimizes for the best-documented 32 rather than returning everything.

The generator was made for the Wikidata-Workshop 2019 in Ulm by bleeptrack and blinry, and its source code is on GitHub for suggestions or bug reports.