Why Globdrop Doesn't Use the Dymaxion Map

An earlier version of this post said Globdrop draws the world on Buckminster Fuller’s Dymaxion projection, from vector data, with hit-testing done on each triangular face. None of that was true. Globdrop has never had a Dymaxion map. Every map in the game is a single image with pins placed on it, and the world scenarios use the Miller cylindrical projection.

The question behind that post is still a fair one, though. The Dymaxion map distorts the size of continents far less than Mercator or Miller do. Why don’t we use it? This post answers as a game-design question: what the game asks of a map, and where the Dymaxion map doesn’t deliver it.

What the Dymaxion map is

Fuller published a first version in Life magazine in 1943, projected onto a cuboctahedron. In 1954 he and the cartographer Shoji Sadao made the version most people know, the Airocean World Map. It projects the globe onto the twenty triangular faces of an icosahedron and unfolds them flat. The cuts between faces run through the oceans, so the continents appear as one almost continuous island (Dymaxion map, Wikipedia).

It is neither conformal nor equal-area. Compared with Mercator it distorts relative sizes less, and compared with Gall-Peters it distorts shapes less. It also has no preferred orientation. Fuller argued that the universe has no “up” or “down”, only “in” and “out”, and the map has no north at the top.

Those are real strengths for some uses. Four features of Globdrop work against them.

1. Players carry a north-up map in their heads

Globdrop asks you to find a place from memory. You see a name, say Ulaanbaatar or the Sabaody Archipelago, and you click where you think it is. The game tests the map you already carry around, and for most players that map has north at the top.

That mental map isn’t a photograph of an atlas either. Barbara Tversky found that memory turns shapes toward the north–south and east–west axes. When she asked Stanford students for compass directions between cities around San Francisco Bay, most answered as if the bay and the coast ran due north–south. Both actually run from north-west to south-east (Tversky, 1981, Cognitive Psychology). Compass directions are part of how people remember where places are.

A Dymaxion net turns each continent to a different angle. Put a player in front of one and the first job is to work out which way Asia is facing, before they can think about where Ulaanbaatar is. That makes the round a test of reading the projection. We want it to test where things are.

Every coordinate system in Globdrop’s code assumes the same thing: x grows to the east and y grows to the south. A scenario’s bounding box is just two corners, topLeft and bottomRight. That’s enough for Miller, UTM, the two equal-area projections and the fictional maps. It isn’t enough for a map where each face has its own rotation.

2. The score is in kilometres

Each round is scored by how far your click lands from the target, in kilometres or miles, against a threshold set per scenario. It’s 2,000 km on the world maps, 500 km for Capitals of Europe and 50 km for the municipalities of Catalonia. Within the threshold, the points fall in a straight line with distance. The how-to-play guide has the full formula.

To make that distance mean something, the game has to measure it the same way everywhere on the map. Today it does so in one of two ways, both in src/engine/coordinate-math.ts:

• On the maps of the real world, it turns your click back into a latitude and longitude with the inverse of the map’s projection and computes the great-circle distance (haversine) to the target. Until September 2026 only Ancient Wonders did this.
• On Catalonia and the fictional maps, it measures the straight line between the two points in pixels and multiplies by a single scale: the real width of the map divided by its width in pixels.

Neither method survives a Dymaxion net as it stands. The pixel method only works because the image is continuous. On a Dymaxion net, two points either side of a cut can be a few hundred kilometres apart on the globe and half the image apart on the screen. Two points on neighbouring faces can also be close on the screen while the straight line between them crosses a gap that doesn’t exist on Earth. A pixel distance on that image has no fixed relationship to a distance in kilometres.

The great-circle method would work, but it needs an inverse projection, and a Dymaxion net needs one for each of its twenty faces plus the logic to know which face a click fell on. That’s a real engineering job, and it wouldn’t fix the next problem.

3. The feedback is a straight line on the image

After every click, Globdrop drops a pin where you clicked and another at the target, and draws a dashed line between them. The top bar shows the distance. We explain why in Designing Globdrop’s feedback loop. In short: the number tells you how far off you were, and the line tells you which way.

On an interrupted map, that line can lie. A miss from Alaska toward eastern Russia could be drawn across half the net, through faces that have nothing to do with the Bering Strait. The distance in the top bar would say “300 km” while the line on screen says “the other side of the world”. Feedback that contradicts itself is worse than no feedback. It teaches the player the projection’s cuts instead of the geography.

4. The distortion depends on the face, not the latitude

On a cylindrical map, distortion depends only on latitude. That makes it predictable: everything at 60° N is inflated by the same amount, and a player can learn to expect it. On a Dymaxion map, distortion depends on where a point falls inside its triangle. Two cities on the same parallel can land on different faces with different distortion.

That’s also why our map distortion calculator has no Dymaxion column. An earlier version of the calculator had one, with numbers from a made-up formula. The note “Why there is no Dymaxion column” on that page explains why one value per latitude would be wrong by construction.

What we give up

Keeping Miller has a cost, and it isn’t small. Miller inflates high latitudes: about 3× in area at 60° and about 6× at 72°. Because Globdrop measures distance on the image with one scale, an east–west miss near Stockholm counts about twice as much as the same miss near the equator. How the Mercator projection changes the size of the world has the numbers.

Where that matters most, we use a regional map instead of the world map. Capitals of Europe is on a Lambert azimuthal equal-area projection, US States on an Albers equal-area conic, and Catalonia on UTM, in metres. Each of these is north-up and continuous, so measuring on the image gives a consistent answer, and each distorts its area much less than the world map distorts the far north. Six coordinate systems in one geography game goes through all of them.

The Dymaxion map is good at showing how the continents connect across the oceans. A game about that would be a different game from Globdrop. Globdrop asks where a place is, on a map you already know how to read, and scores the answer in kilometres. For that, a north-up map with straight meridians is the better fit.