The leopard complex (appaloosa)

What it does

Crossing two of them

Its gene carrot

Gene keys

horsegenetics.leopard, horsegenetics.patn1, horsegenetics.patn2

Alleles

LP lp  ·  PATN1 n  ·  PATN2 n

Outcomes (LP)

wild, mottled, varnish-roan, leopard, fewspot, blanket, snowcap, semi-leopard

Outcomes (PATN1/PATN2)

wild only — pure modifiers, they never paint

Priority

leopard 73 (white band), PATN1 93, PATN2 94 (paints-nothing band)

Default alleles

lp, n, n

Wild frequency

LP ~1 in 40 per allele; PATN only drawn onto an LP founder

Deterministic

no — every visible LP outcome varies per horse

Epi draws

off the expressing LP copy: nextLong() seed + 2–3 nextFloat() jitters per painter

Health

LP/LP → CSNB (night blindness), informational — costs nothing

Reads

coatDependsOn() = [patn1, patn2]

The outcome grid

The visible pattern is a function of this locus's zygosity and the two modifier pairs. expressionOf(pair) — which only sees the LP pair — can only give a coarse answer (varnish-roan for any LP horse, which keeps the gallery to two leopard-complex pens). The pipeline calls expressionIn(pair, genotype), which reads PATN1 and PATN2 and returns the real one of eight:

• mottled — light, uneven roaning that keeps its colour over the bones, a few small white spots, plus the always-on triad. The appaloosa “look” without a bold pattern.
• varnish-roan — heavy, uneven whitening that reaches the head and leaves darker varnish marks over the horse’s bony prominences, so it ends up outlined by its own anatomy. Not true roan, and not a stronger version of it.
• leopard — a white body carrying round dark spots of the base colour, head to hoof.
• fewspot — almost solid white, a handful of spots left.
• blanket — a solid white sheet over the hips with dark spots inside it; the rest of the horse its ordinary colour, legs coloured.
• snowcap — the same blanket with no spots.
• semi-leopard — a spotted blanket that reaches forward over the barrel, between blanket and full leopard.

The painters

All phase 1, like the other white genes — they only ever push pigment toward zero, so the bald white template shows through. Every number comes from the expressing LP copy's stored values, so a foal that inherits the copy reproduces the exact pattern.

All four go through one private whiten(f, px, py, amount), which is PigmentField.whiten and nothing else. That matters most for paintVarnish, which is all partial whitening: on a black-based appaloosa the old per-gene approximation left a side view carrying 211 warm texels — the varnish reading tan rather than grey — against the 64 a plain black horse gets from the template's own shading. It is 59 now.

Varnish marks — the horse outlined by its own skeleton

A varnish horse whitens around its bones without erasing them. Pigment persists over the cheekbones, shoulder blades, elbows, hips, stifles and knees, and those residual darker zones are what the pattern is named for. So paintVarnish subtracts a retention field before it roans anything:

Every mark is derived from the mesh’s own part boxes rather than written down as coordinates, so a mesh change moves them with it. The strongest mark wins rather than the sum — two overlapping marks must not add up to a solid patch — and the distance is wobbled by a noise field so the island’s outline is ragged. A real varnish mark follows the bone loosely, not exactly. How deeply the marks hold is rolled per horse, because “varnishing out” is a process and different horses are at different points along it.

Against classic roan

RoanPatternsTest pins the comparison through the real pipeline, and it is the reference’s own identification checklist:

“White finds white” is not applied here yet — leopard composes with the other white loci at priority 73 (early), so a leopard horse that is also sabino simply gets sabino white added on top. Multi-pattern appaloosas want an in-game look; flagged in verification.

The always-on triad

Every LP horse — any non-wild outcome — also gets, in the overlay phase (so the eye redraw cannot wipe them):

• Striped hooves — vertical light/dark bands over the bottom HOOF_FRACTION of each leg, from a body-space phase; the light bands blend toward a pale horn colour.
• A white-rimmed eye — shadeEyes toward a cool near-white. Because the shade is luma-scaled, the bright sclera lifts while the dark pupil stays dark.

Mottled skin — the third member of the triad — is not drawn; the coat has no skin layer, only hair. Flagged in the roadmap.

CSNB

Every LP/LP horse has congenital stationary night blindness — the same TRPM1 change that makes the pattern also stops the retina resetting in the dark. The mod has no night vision for a horse to lose, so like deafness on a splash homozygote it is informational: shown on the info panel and the paper dump, costing the horse nothing. One copy carries the pattern with none of the defect, which is why the pairing worth breeding for is also the one that hides it.

In the wild, and in breeds

LP founders are ~1 in 40 per allele — a bit under 5% carriers, well under 1% LP/LP — and a wild founder only draws a variant PATN copy if it already rolled LP at LeopardGene (which sorts earlier, so the founder context has it). So a wild spotted appaloosa is a genuine find; the pattern is common once you breed for it.

Appaloosa, Knabstrupper, Kiger Mustang, Mustang and Shetland Pony roll real LP + PATN pools — the Appaloosa and Knabstrupper heavily, so most of a herd shows a pattern; the mustangs and the Shetland lightly.

Why the modifiers are separate genes

LP is one locus (TRPM1). PATN1 and PATN2 are real, independently inherited modifier loci, so they have to be separate genes — a breeder works toward “this line carries PATN1” the same way they work toward any other allele. They just have no phenotype of their own: a PATN1 carrier with no LP looks like any other horse.

That creates a texture-cache problem. A gene is left out of a horse's coat code unless affectsCoat() is true, and every one of PATN's own combinations is a wild type — so a LP/LP PATN1/PATN1 horse and a plain LP/LP one would collide in the cache and render identically. The fix is a new hook, Gene.coatDependsOn(): LeopardGene names both modifiers, and Genotype.coatCode() folds their alleles into the texture key — but only for a horse that actually carries LP, so a PATN carrier with no leopard complex still shares its texture. BreedFounder likewise treats a named modifier like a coat gene (forced wild unless the breed lists it).