Particle

Crossing two of them

It paints nothing

All 87 outcomes are wild types, so affectsCoat() is false, the locus is out of the texture key, and GenotypeCatalog collapses all 861 combinations to a single entry — forty alleles for no catalogue growth at all, the trick sex, milk and verdant already use.

Unlike those two, though, there is no visibility gap here: the whole gene is the thing you can see. It is the answer to the complaint milk and verdant both attract — an effect a player can watch without having to test for it.

Its gene carrot

Gene key

horsegenetics.particle

Priority

150 (magical band)

Alleles

40 variants + n — see the table

Combinations

861, all carryable — 87 outcomes

Outcomes

wild, 40 single, 46 codominant double

Shows when

neither copy is n — see below

Default allele

n

Wild frequency

0.15% per homozygote, 0.04% per codominant pair — about 7.8% of founders trail something, and no founder is a carrier

Founder draws

1 × nextFloat()

Coat effect

none, ever

Emission

while moving on the ground, 20% of ticks, 1–4 particles a firing

One copy of n and the locus is silent

The locus asks for agreement before it draws anything at all. Two copies of the same allele, or two alleles of one family — and nothing else. Two variant alleles that are not codominant partners are simply two carriers: a Dst/Soul horse trails nothing, and passes both on.

Declaration order survives, but it is no longer a dominance rank. The alleles are declared in it because AllelePair canonicalises on Allele.order(), and that is what makes copy(0) and copy(1) address the two halves of a codominant pair in a fixed order — which matters, because each half draws its own colour and its own place on the horse.

n is an off switch and it beats everything — and now so is any allele without its partner beside it.

Codominance is by family

Every allele belongs to at most one family, and two different alleles of one family both show at once — in two different colours, from two different places on the horse. Nine families cover twenty-nine of the forty alleles and produce the locus’s 46 double outcomes; the other eleven alleles belong to no family and never stack with anything.

The flames and the smokes are one family of eight, not two of four. Any -flm stacks with any -smk as well as with its own kind, which is 28 of the 46 doubles all by itself — get that grouping wrong and most of the locus’s double outcomes quietly disappear. A unit test pins it.

Two alleles whose tokens look related but are not: Dstrn (enchanting glyphs) is not a dst, and Lmstr (totem sparks) is not a str. Both are ungrouped, on purpose — the family is about what the particle is, not what the token spells.

Everything you can see about it is epigenetic

The allele names a particle and nothing else. Everything else is stored on the allele copy that carries it and read by name:

Every value is stored whether or not this particular particle uses it, so a variant that starts or stops caring about its second colour costs nothing and moves nothing. This used to be a fixed draw order, and the particle locus is the gene that paid for that lesson: the position of a draw was the meaning of each number, so a conditional draw quietly changed every horse in every save. Values are read by name now and the hazard is gone.

So two horses that are both Rflm/Rflm are not the same horse: one trails four red flames off its front hooves, the other one green flame off its tail. And because each copy of a codominant pair carries its own draw, the two halves are independent — red flames off the front hooves and blue smoke off the tail is one horse, and nobody wrote that combination down.

A foal that inherits the copy inherits the exact number, the way magic body size and mane colour do. That is what makes a particular look worth breeding from rather than worth re-rolling.

The machinery it needed

Abilities used to be a pure function of the genotype (AbilityContribution), which is right for every gene where the alleles fix the behaviour — two Hlr/Hlr horses heal identically, and should. This locus is the case that cannot express, so it is the first EpigeneticAbilityContribution: the same interface one layer along, handed an GeneEpigenetics so it can read copy(0) and copy(1) separately. It is the exact twin of what EpigeneticTraitContribution did for the body, and GeneEpigenetics moved from common/trait/ to common/genetics/ so both sides could use it without the two packages depending on each other.

HorseAbilities.activeFor gained an Epigenome overload, and the translator’s per-horse cache is now keyed on both code strings — keyed on the genotype alone it would have handed one horse another horse’s colours.

The emitter verb grew with it: a second colour, a count, a spare data number, and five new body sites beside the four single-point anchors it had. All of it is usable from a gene file today, which is the property the shared vocabulary was chosen for.

Every wild horse at this locus is one you can see

The founder table is not the random-mating shape, and that is deliberate. It lists the forty homozygotes at 0.15% each and the forty-six codominant pairs at 0.04% each, and nothing else. No X/n carrier. No cross-family Dst/Bflm quietly sitting on a soul flame nobody can see.

In aggregate that is about 7.8% of wild horses trailing something, while any one named particle is roughly one horse in six hundred and sixty and any one double is one in twenty-five hundred. Meeting a particle horse is a good day; meeting the one you were looking for is a find.

The forty alleles

In declaration order. That order fixes which half of a codominant pair is copy(0); it is not a dominance rank, and where two alleles are not in the same family neither of them shows.

Ten of them are not the particle they were named after

The list this gene was designed from uses Bedrock particle names, and ten of them have no Java equivalent in 26.1.2. Each was mapped to the nearest real particle, and the substitutions are recorded here because the allele tokens still carry the original intent:

Two alleles were cut rather than substituted. Bigflm (mobflame_emitter) and Cndlflm (candle) both land on a particle another allele already has — flame and small_flame — and Java flames are SimpleParticleTypes carrying neither a colour nor a scale, so there would have been no way to tell the pairs apart. Two alleles a player cannot tell apart are worse than two alleles that do not exist, so the list is forty rather than forty-two.

Why one locus and not forty genes

Forty independent two-allele genes would let a horse carry all forty at once, and then the interesting question — which of these does this horse trail — stops having an answer. Every serious line converges on a horse emitting everything, and the fortieth gene is worth nothing because the horse was already covered in particles.

One locus with forty alleles says the opposite, and says it structurally rather than by a rule anyone has to enforce. A horse has two copies of chromosome whatever-this-is, so it can carry two of these and no more; a third is not rare, it is impossible. It is exactly the argument that folded dominant white and sabino into KIT, at forty times the scale.