Document GP-DOC-03: Genetics

The Genome

Every specimen is defined by a genome of 12 loci, each carrying 4 possible alleles (index 0-3). Every locus is diploid: it stores two alleles, one from each parent. The haploid space alone is 4^12 ≈ 16.7M distinct genotypes; the diploid space is far larger. The design goal is a space no human resolves by hand or in a spreadsheet, while staying fully deterministic, the legal and competitive backbone of the game.

The genome is stored on-chain, in the asset's attributes, as the carried genotype (both alleles per locus), not merely what is expressed. Nothing is hidden. There are no money-gated "hidden IVs". The Specimen Card renders the genome in full; transparency is the aesthetic.

The twelve loci

#	Gene	Key	Allele `0` → `3`	Expression
1	Limb count	`LMB`	2 / 3 / 4 / 6 legs	Dominance
2	Limb length	`LEN`	short → long	co-dominant (mean)
3	Limb girth	`GIR`	thin → thick	co-dominant
4	Torso mass	`MAS`	light → heavy	co-dominant
5	Torso length	`TOR`	compact → elongated	co-dominant
6	Stance (center of mass)	`STA`	low → high	co-dominant
7	Joint tuning	`JNT`	supple/elastic → rigid	co-dominant
8	Fiber type	`FIB`	slow (endurance) → fast (explosive)	co-dominant
9	Muscle power	`PWR`	weak → strong	co-dominant
10	Tendon	`TEN`	stiff → elastic	co-dominant
11	Metabolism	`MET`	low → high (energy capacity)	co-dominant
12	Foot type	`FUT`	hoof / pad / claw / web	Dominance

Ten loci are quantitative (co-dominant). Two, LMB and FUT, are categorical and resolve by dominance. This split makes limb count and foot type the two most "visible" traits on the card, and the source of emergent marketplace rarity.

Quantitative loci: the co-dominant mean

For the ten quantitative genes, the expressed value is the mean of the two alleles, normalized to [0,1]:

g = (allele_a + allele_b) / 2 / 3     // ∈ [0,1]

A locus [0,0] expresses 0.0; [3,3] expresses 1.0; [1,2] expresses 0.5.

Categorical loci: dominance and carried recessives

LMB and FUT do not average. Each allele maps to a category; the dominant of the two is expressed while the recessive is carried (and recorded on-chain):

LMB alleles → 2, 3, 4, 6 legs        dominance: 4 > 3 > 2 > 6
FUT alleles → hoof, pad, claw, web   dominance: claw > pad > hoof > web

The most recessive alleles, LMB=6 (hexapod) and FUT=web, surface only when both alleles at the locus rank at or below them. A six-legged creature appears only when both parents carry the 6; webbed feet only when both carry web. Rarity is therefore emergent: a property of the dominance lattice and the breeding rules (see Breeding), not a tier stamped on at mint.

Genome → morphology

The expressed genome is turned into four morphological intermediates, from exact engine code:

mass       = 0.25 + 0.75*MAS + 0.15*GIR + 0.04*(legs - 2)     // ~0.40 .. 1.40
force      = 0.15 + 0.80*PWR + 0.35*GIR                        // ~0.15 .. 1.30
ptw        = force / mass                                       // power-to-weight
foot_grip  = { hoof: 0.60, pad: 0.85, claw: 1.00, web: 0.70 }[FUT]

legs is the expressed limb count (2, 3, 4, or 6); gene keys (MAS, GIR, PWR, …) are the normalized [0,1] values above.

Leg factors: the core of the epistasis

Limb count rewires what every other gene does. Each expressed leg count applies a fixed vector of multipliers to the derived stats, and this table is where the archetypes come from:

legs	speed	accel	stability	agility	efficiency
2	1.08	0.90	0.20	1.05	1.10
3	1.00	1.00	0.55	1.10	1.00
4	0.98	1.08	0.85	0.95	0.92
6	0.90	1.18	1.00	0.75	0.80

The biped is a fast but unstable sprinter with no recovery; the tripod is the agility peak, the "freak" that ice and technical circuits reward; the quadruped is balanced; the hexapod is a stable, powerful grinder that is slow, clumsy in turns, and energy-hungry.

Why it does not resolve in a spreadsheet

Depth does not come from the gene count. It comes from three structural properties encoded above:

Pleiotropy: one gene moves several stats at once (GIR feeds both mass and force; FIB trades speed against stamina).
Epistasis: genes rewire each other; the leg-factor table changes the meaning of every quantitative gene.
Trade-offs: there is no stat budget coded anywhere; the tensions in the formulas mean no genome maxes everything.

Combined with the hourly biome rotation (see Terrain Matrix), there is no global optimum, only optima conditional on the circuit.

Carried versus expressed

The genome tracks both alleles at every locus, so the full genotype (including recessives that never surface in the phenotype) is public and travels with the asset. This is what makes a cross fully previewable before payment and the pedigree verifiable (see Breeding). The engine reads the genome the same way in the browser and in the headless balancing harness: one source of truth, no separate model.

Reserved for later

Four traits are documented but not implemented in v0: a tail/counterweight gene, split fore/hind joint tuning, thermoregulation, and a cosmetic foil gene. They are mutation-gated: reachable only through catalysts obtained by playing (see Breeding), never sold, never present in a fresh wild genome.

Next: Stats & Morphology, the eight stats these morphological values drive.