Borrowing an ocean: cutting a game world from real survey data

Every fishing game has a seafloor. Almost none of them have a real one — and in a game where players read the bottom for a living, the difference is everything. Anchor Point's world isn't sculpted; it's surveyed. This is how one developer cut 64 × 64 kilometres of Alaska from real GEBCO bathymetry data, turned it into an Unreal Engine 5 world, and why borrowing the planet was the best level-design decision I never had to make.

This is the third entry in the Anchor Point Journal. The first post covered how a solo dev scopes an ambitious game at all; the second went deep on the weather engine that runs the sky and the market. This one is about the ground under all of it — because weather only matters if the water it moves across is worth reading.

You can't fake a seafloor players stare at all day

In most games, terrain is scenery. In a commercial fishing sim, the seafloor is the content: the player spends whole sessions watching a sounder trace it, guessing what lives on it, betting fuel money on being right. Hand-authored underwater terrain has tells — the too-even slope, the convenient channel, the drop-off placed where a designer wanted a "cool moment". Players who fish learn to smell it. The bottom reads like a level, and the skill of reading nature collapses into the skill of reading the designer.

So I refused to sculpt it. Anchor Point's seafloor comes straight from GEBCO — the global bathymetric survey assembled from real soundings. The shelves, the channels and the thousand-metre drop-offs are where the survey says they are. My home coast around 59.77°N, 151.83°W was carved by actual glaciers over tens of thousands of years, and it shows: nothing about it is even, convenient, or placed for a moment. It's better level design than I will ever do, because nobody designed it.

From survey grid to Unreal Engine 5 landscape

The pipeline is deliberately boring, which is the point — a solo dev's pipeline has to be. GEBCO publishes the ocean as a grid of depth values. I crop a window around the home coast, resample it to a heightmap, and bring it into Unreal Engine 5 as the landscape the water sits on. Then comes the one step that actually matters: calibration. The vertical scale is tuned so that depths in the game are the depths in the survey — when the chart says the wall bottoms out at 1,025 metres, the sounder agrees, because they are reading the same numbers.

That's the rule the whole world is built on: one source of truth for depth. The same grid that shapes the terrain also drives the 128 × 128 navigation grid the server thinks in — 16,384 cells at 500-metre resolution — and the sounder the player squints at in a squall. Rendering, simulation and instruments all read from the same chart. The sounder tells the truth because there is only one truth to tell.

What the real seafloor gives you for free

Here's what surprised me: I imported terrain, and gameplay fell out of it.

• Ten grounds with real character. The zones aren't rectangles on a design doc — they're computed from depth, distance from shore and channel detection, and their borders follow the actual coastline. The estuary at the river mouth really is sheltered; the outer grounds really are exposed.
• Channels that funnel everything. The same trench that funnels current funnels the weather. When a front rolls down the inlet, the geography decides who feels it first — not a script.
• A shelf break where the money lives. Where the depth lines crowd together, the big halibut hold. That's not my invention; it's how the real coast fishes. Learn the chart and you've learned something true.
• An endgame written in rock. The progression from the estuary, across the shelf, down the slope to the wall at the ice edge is an economic ladder the terrain built for me: deeper water, bigger fish, worse weather, longer runs home.

Where the data ends and the design begins

Borrowing the planet doesn't mean abdicating design — it means moving it. The design act isn't sculpting the terrain; it's choosing the window. I spent longer picking those 64 kilometres than I would have spent sculpting them: the frame had to contain a sheltered estuary for the first boats, a working shelf for the middle game, and a genuine abyss for the endgame, all within a sail of six harbours and a home port.

On top of the fixed terrain sits the layer that keeps it alive: 416 plotted fishing spots, of which roughly 250 are live in any given week, replotted every Monday. The seafloor never lies, but the fish move across it — so the chart stays true while the fishing stays fresh. And because discovery is persistent fog-of-war, every captain's chart fills in differently. Your chart is your biography.

The last design layer is the ecology reading the same data. Twenty-six species carry real depth ranges, ground preferences and weather moods — herring don't care what I want, they care what the water is doing. The terrain gives the stage its shape; the weather engine decides the play. Two systems, both anchored to the same truth, multiplying each other. More on the world & ecology →

Surveying my own world

Real terrain has one drawback: nobody balanced it for you. A hand-built map comes pre-flattened toward fairness; a real coast doesn't owe anyone a living. So the same Monte Carlo harness that tunes the weather runs thousands of automated careers across every ground, boat and season, and reports which zones quietly starve a captain and which ones print money. Where the numbers break, I adjust the economy around the terrain — never the terrain itself. The chart stays honest; the prices do the balancing.

What I'd tell another solo dev building a world

• If your game touches the real world, there's a dataset for it. Bathymetry, elevation, tides, traffic — open data is a level-design team you don't have to hire.
• Curate a window, don't generate a planet. Picking the right 64 kilometres is the design. Infinite procedural oceans are all the same ocean.
• Keep one source of truth. Terrain, simulation and instruments should read the same numbers. The moment they diverge, your world starts lying.
• Let real structure create local knowledge. If the geography is true, "knowing the waters" becomes a genuine player skill — the rarest kind of content.
• Balance around the land, not by flattening it. Tune prices, spots and schedules. Never sand the interesting parts off the world to make it fair.

The deepest compliment this world could ever get is a player pulling up a real chart of Kachemak Bay and recognising the game in it. That's the bet: authenticity isn't a texture you paint on at the end — it's a decision you make at the very first import.