Give the island preset a main landmass worth landing on #161

New issue

Open

opened 2026-10-11 20:23:46 +01:00 by liamjd · 0 comments

liamjd commented

2026-10-11 20:23:46 +01:00

Owner

Milestone 2, terrain goal: "Islands map has more land". Also resolves the open decision recorded in #16 ("The island preset nearly drowns at small grid sizes").

Problem

The island is mostly sea at every size, and almost none of its land is buildable:

size	walkable	main landmass	flat 3×3 rocket sites (min / avg)	seeds with no site
16	23 %	12 %	0 / 0	10 / 10
32	23 %	19 %	0 / 0.1	9 / 10
64	24 %	19 %	1 / 14	0 / 10
128	23 %	19 %	25 / 83	0 / 10

(10 seeds each.) Three things combine:

The noise is centred on the datum, which is also the water line, so roughly half of the tiles start under water before the falloff touches them.
falloff_start = 0.80 then drags the outer ring down to the seabed.
The noise breaks the land into several islands, and only the largest counts as the main region.

Proposal

Raise the island's land relative to the water. Either bias the noise upward before the falloff, or lower the water line (#164 makes the water level its own setting, which is the cleaner lever). The target is a main landmass of 45–60 % of the map, with a clear coastline all round. Treat the number as a tuning target, not a law.
Bias toward one landmass. One way is a radial dome blended under the noise (lerp(n, dome, k)), so the centre is reliably land and the coast is where the noise wanders.
With #159 in place, size dependence should largely go; confirm it with the probe rather than adding size corrections.

Acceptance criteria

Main-region share is 45–60 % at NORMAL, averaged over 10 seeds, and the minimum seed stays above ~35 %.
At TINY (#158) the island keeps a main landmass of at least ~35 %. That answers #16's question: small islands are supported, not unsupported.
Every seed sampled has at least one flat 3×3 site on the main landmass, even before #163 guarantees one.
The border ring is still water: it should still read as an island.
#16's "island: the border is below the datum, the interior is not" shape check is added here, since this ticket is what makes it true.
Headed screenshots at TINY and NORMAL.

Depends on

#159. Works best after #164, if the water level is the lever chosen.

Milestone 2, terrain goal: "Islands map has more land". Also resolves the open decision recorded in #16 ("The island preset nearly drowns at small grid sizes"). ## Problem The island is mostly sea at every size, and almost none of its land is buildable: | size | walkable | main landmass | flat 3×3 rocket sites (min / avg) | seeds with no site | | --- | --- | --- | --- | --- | | 16 | 23 % | 12 % | 0 / 0 | 10 / 10 | | 32 | 23 % | 19 % | 0 / 0.1 | 9 / 10 | | 64 | 24 % | 19 % | 1 / 14 | 0 / 10 | | 128 | 23 % | 19 % | 25 / 83 | 0 / 10 | (10 seeds each.) Three things combine: - The noise is centred on the datum, which is also the water line, so roughly half of the tiles start under water before the falloff touches them. - `falloff_start = 0.80` then drags the outer ring down to the seabed. - The noise breaks the land into several islands, and only the largest counts as the main region. ## Proposal - Raise the island's land relative to the water. Either bias the noise upward before the falloff, or lower the water line (#164 makes the water level its own setting, which is the cleaner lever). The target is a **main landmass of 45–60 % of the map**, with a clear coastline all round. Treat the number as a tuning target, not a law. - Bias toward **one** landmass. One way is a radial dome blended under the noise (`lerp(n, dome, k)`), so the centre is reliably land and the coast is where the noise wanders. - With #159 in place, size dependence should largely go; confirm it with the probe rather than adding size corrections. ## Acceptance criteria - [ ] Main-region share is 45–60 % at NORMAL, averaged over 10 seeds, and the minimum seed stays above ~35 %. - [ ] At TINY (#158) the island keeps a main landmass of at least ~35 %. That answers #16's question: small islands are supported, not unsupported. - [ ] Every seed sampled has at least one flat 3×3 site on the main landmass, even before #163 guarantees one. - [ ] The border ring is still water: it should still read as an island. - [ ] #16's "island: the border is below the datum, the interior is not" shape check is added here, since this ticket is what makes it true. - [ ] Headed screenshots at TINY and NORMAL. ## Depends on #159. Works best after #164, if the water level is the lever chosen.

liamjd added this to the Terrain Generation milestone

2026-10-11 20:23:46 +01:00

liamjd added the

enhancement

label

2026-10-11 20:23:46 +01:00

liamjd referenced this issue

2026-10-11 20:23:48 +01:00

Guarantee a flat landing site for the colony rocket #163

liamjd referenced this issue

2026-10-11 20:28:55 +01:00

Terrain statistics probe for milestone 2 tuning (#159) #168