About
An independent educational project
A visual judgment game built from open museum collections.
Overview
VBI · QVANDO · QVIS asks you to place, date, and attribute five Old Master paintings before the museum label is revealed. The current edition draws from 6,415 paintings from ten public collections. Each answer is scored against a reviewed, reproducible dataset and followed by the catalogue record and available production evidence.
The game
Look first. Then place, date, and attribute.
A game is a five-painting encounter. The title, artist, date, and museum remain out of view until all three judgments are committed.
- I · VBIGeography3,000
Place the painting at the most precise level its evidence supports: city, region, or country.
- II · QVANDODate3,000
Estimate a year. The accepted catalogue interval earns full credit; points then fall with distance.
- III · QVISAttribution3,000
Name the artist or school while respecting qualifiers such as workshop of, circle of, or after.
Each judgment is worth 3,000 points: 9,000 per painting and 45,000 per game. Date scoring measures distance from the accepted interval. Attribution scoring recognizes common name forms, modest spelling errors, and qualified relationships. The reveal explains the score, shows accepted alternatives, and links back to the museum record.
Five works are drawn from the full catalogue. The server calculates every score, and a chosen public username can appear on the leaderboard.
Build a set by period, school, region, attribution type, or museum collection without affecting ranked results.
Current edition
Ten collections, one reviewed corpus.
The game uses fixed snapshots of the participating museums’ catalogues rather than allowing a live catalogue change to rewrite a round in progress.
These museums were chosen because they publish reusable collection data and images at meaningful scale, retain stable object identities and catalogue links, and offer complementary European painting holdings. Together they also test the same research method across different APIs, vocabularies, and cataloguing traditions without weakening the project's rights or evidence standards.
- The Metropolitan Museum of Art1,647
- Cleveland Museum of Art388
- Art Institute of Chicago540
- National Gallery of Art1,004
- SMK — National Gallery of Denmark757
- The J. Paul Getty Museum353
- Yale University Art Gallery406
- Städel Museum422
- Minneapolis Institute of Art233
- Nationalmuseum, Stockholm665
Rights, image, object-type, date, European-scope, and geography rules are applied collection by collection. Qualified attributions remain qualified, ambiguous records are excluded, and closely related works are prevented from appearing together too frequently. Admissions, exclusions, sources, and transformations remain auditable, and every published work has complete, scorable geography.
Method
Preserve museum identifiers, rights information, catalogue links, images, and the fields needed to reconstruct each selection.
Apply a published collection policy, retain qualified attributions, and record admissions and exclusions in auditable ledgers.
Compare catalogue evidence with matched identities, dated artist activity, and object-specific research without turning nationality into production place.
Generate compact runtime data and scoring geometry from hash-checked snapshots, then rebuild and test the edition before release.
Reproducibility
Source captures, normalized records, research decisions, boundaries, and generated gameplay files carry recorded hashes. Automated checks rebuild the geography and scoring artifacts before release.
Geography
Production place, at evidence-supported precision.
Nationality, birthplace, travel, subject matter, provenance, and place of death may be relevant context, but none independently establishes where a painting was made.
The resolver compares museum records with matched artist identities, dated RKD activity, selected Wikidata alignment, and object- or series-specific scholarship. A more precise answer is required only when every full-credit alternative supports it; genuine alternatives can all receive credit.
Gameplay uses present-day labels and generalized modern boundaries as a consistent scoring instrument. They are not claims about historical sovereignty, cultural identity, or the exact extent of an early modern polity.
Systems
Stable play and inspectable results.
The application preserves source evidence, makes transformations explicit, and keeps authoritative scoring on the server.
The React interface is built with Vinext and Vite and served by a Cloudflare Worker that also runs the game APIs. Every painting is one of the game's own JPEG derivatives of the museums' public-domain images, generated once at each museum's published size and stored by the game, so no request while playing fetches from or transforms a museum's image.
Cloudflare D1 and Drizzle store ranked games, the leaderboard, accounts, and optional cross-device progress. Guest Practice remains device-local.
Signed game tokens, immutable first submissions, and server-side reconstruction keep accepted answers and scores outside browser authority.
Leaflet and MapLibre render the visible map; versioned Natural Earth geometry supplies stable gameplay boundaries.
The visible map uses Leaflet, MapLibre, OpenFreeMap, OpenMapTiles, OpenStreetMap data, and Natural Earth relief. Separate versioned Natural Earth geometry controls scoring, so a visual tile change cannot alter an answer. Optional Google sign-in carries progress between devices while retaining only Google's account-specific identifier; the public identity remains the username chosen in the game. See the privacy policy for details.
Future work
Where the project goes next.
The next phase is deeper research, stronger study tools, and careful growth without lowering the evidentiary standard.
Expand object-specific production research and make uncertainty, competing evidence, and accepted alternatives clearer in every reveal.
Develop Practice into a stronger curriculum tool with saved sets, comparisons, and clearer progress through periods, schools, and collections.
Explore time-aware cartography without confusing historical borders with settled facts or changing the fairness of the scoring contract.
Add museums one at a time, prioritizing openly reusable European painting collections with stable images, durable identifiers, structured catalogue evidence, and terms that support reproducible offline snapshots. Each enters only after provider-specific reconciliation and geography review.
Proposed research remains outside gameplay until it is reviewed, accepted, regenerated, and tested for a planned release. Data and scoring geometry carry versions and hashes, and completed rounds are not retroactively recalculated. A further release-hardening step is to bind each newly issued game to the exact artifact hashes used to score it.
Acknowledgements
Made possible by public collections.
VBI · QVANDO · QVIS is an independent educational project and is not affiliated with, operated by, or endorsed by The Metropolitan Museum of Art, the Cleveland Museum of Art, the Art Institute of Chicago, the National Gallery of Art, SMK, the J. Paul Getty Museum or Getty Trust, Yale University or the Yale University Art Gallery, the Städel Museum, the Minneapolis Institute of Art, Nationalmuseum (Sweden), the Wikimedia Foundation, RKD, or any of the mapping and infrastructure providers whose resources it uses. Their open-access initiatives make the game possible, and their catalogue records remain the primary references for the works included. The project was initially inspired by Anthropeum, a game centered on objects from The Metropolitan Museum of Art’s collection, but has evolved substantially over the course of its development. We are grateful to the many friends and colleagues who have tested the game, offered feedback, and provided corrections. We welcome further corrections and look forward to continuing to improve and expand the project.
Suggest a correction