This one isn't a feature. It's a fix for a tiny rendering quirk that has been quietly there since 2024, and a piece of cleanup that makes the next round of polish easier to land.
The terrain in Sheep Dog Sim is a heightfield — a grid of height samples that defines how high the ground is at every (x, z) point on the map. Everything that "sits on the ground" — the dog, the sheep, the trees, the rocks, the camera — needs to ask the heightfield "what's the Y here?" and place itself accordingly.
For a long time, there were two slightly different ways of asking that question: one path that the visible terrain mesh used, and another path that everything else used. The two answers agreed almost everywhere, except in a small set of edge-case pixels where they could disagree by a few centimetres. To compensate, an old workaround added a defensive 5-centimetre lift to one of the paths. It mostly worked — you'd only ever notice it if you put the camera very close to a sheep on a particular slope at a particular angle.
But "mostly works" is the kind of phrase that ends up costing time later. If you're trying to add a new on-ground system — a fence editor, a new prop type, an objective marker — and you can't predict whether the ground is actually at Y or Y + 0.05, every new feature inherits the ambiguity.
The terrain math now lives in one place. There's a single function that knows how to compute the visible-mesh Y at any (x, z) point, and every consumer (the renderer, the prop placement, the camera, the future fence editor) calls it. The compensating 5-centimetre lift has been deleted.
The deterministic simulation — the part of the code that runs both on the server and on every connected client to keep multiplayer in sync — still uses the raw heightfield, not the mesh-aligned one. That separation is intentional. The simulation cares about being identical across machines, byte for byte; the renderer cares about looking right. They're allowed to disagree by a hair, as long as the disagreement is now centralised and understood.
Honestly? Probably nothing. If you spent a lot of time in close-up Follow camera mode chasing a sheep across uneven terrain, there's a chance you'll notice that contact looks a touch tighter. Otherwise this is invisible polish — the kind of cleanup that doesn't sell screenshots but makes the next few feature cycles ship faster and with fewer "wait, why does this float" bugs along the way.
← Back to game