Document GP-DOC-07: Species

The Eight Species

A species in GENOPOOL is not a power class and not a stat set. It is an endemic allele-sampling bias: when a wild specimen of species X is generated in its biome, the draw distribution over the alleles of certain loci is weighted. Expression (morphology → the eight stats) then follows the formulas in The Genome and The Eight Stats, identical for every creature.

Three locked consequences:

1. Never a stat re-roll. The bias acts only upstream, on the draw weight of alleles, never on the computed stats. Two creatures with the same alleles have exactly the same stats, whatever their species of origin. There is no species-linked stat adjustment after the draw.
2. Seeded randomness, never Math.random. The weighted draw derives from the on-chain hourly seed (see Wild Spawns and Provably Fair). Two spawns from the same seed produce an identical genome.
3. Rarity emerges, it is never declared. No species pins a target stat. A Capra bias toward the LMB=6 carrier makes the hexapod less rare in Capra; it does not guarantee it: the hexapod stays recessive. Rarity (tripod, hexapod, terrestrial web-foot) emerges from recessivity plus bias, not from a rarity tag.

The species therefore shifts a profile, never a rating. It gives each "freak" a biome where it is less rare: niche demand on the marketplace, not a single meta.

The bias vector

Each bias is a sampling-weight vector [w0, w1, w2, w3] over the four alleles of a locus. Default is uniform [1,1,1,1]; unlisted loci are uniform (a no-op). The draw is per diploid allele: each locus receives two independent weighted draws (maternal, paternal), and dominance / co-dominant averaging apply afterward, unchanged.

allele indices
  LMB (limb count) : 0 = 2 legs | 1 = 3 legs | 2 = 4 legs | 3 = 6 legs
  FUT (foot type)  : 0 = hoof   | 1 = pad    | 2 = claw   | 3 = web
  co-dominant loci : 0 -> 3 from the low to the high end of the gene

recessivity (dominance order)
  LMB : 4 > 3 > 2 > 6   -> the hexapod (6) is the deepest recessive
  FUT : claw > pad > hoof > web  -> the web-foot is the deepest recessive

A ×N multiplier scales the weight of the targeted allele inside the vector; the distribution is renormalized at sampling time (Σ = 1). The stored vector is never pre-normalized.

The roster

The canon roster is exactly these eight families. For each: Latin genus plus endemic code, biome(s), expected archetype, and the key numeric biases (from SPECIES.md §2).

Values are v0 contractual and tunable by the balancing harness, see Balancing. The contract here is the structure (which locus is biased, in which direction), not the third decimal.

Reading the roster

• Equus is the most conformant family: a balanced hoofed quadruped, few freaks, little market premium. The stable showcase.
• Capra and Myria are the two hexapod-carrier reservoirs (6p ×2 and ×4). Even at ×4, the expressed hexapod stays conditioned on two carrier parents: the rarity holds.
• Aranea and Serpens are two distinct tripod niches (3p ×2 vs ×3), compact vs low-and-elongated, so two different tripod builds trade on the market.
• Rana is the family where the deepest-recessive web-foot is least rare: terrestrial web-foots stay a scarce niche.
• Cursor, the eighth family, gives a face to FLAT terrain (light long-limbed biped). Its bias is a watched item: a global tunable amplifies or attenuates the whole Cursor bias without touching engine code, so FLAT stays winnable by non-Cursor builds.

No active abilities in v0

Decision (v0): no player-triggered active abilities. All species differentiation is 100% passive: allele bias → morphology → the eight stats → the terrain matrix. No species has a button to press during a race. The piloting skill layer is exclusively the one already specified: optimal throttle/effort, the execution coefficient, and energy-budget management (see The Eight Stats).

Rationale: the skill/build model is proven on the passive substrate first, before any second skill axis is layered on. Aranea's advantage on ICE / TURNS / DECLINE is passive (the expressed tripod receives the peak agility factor of the morphology table), not a triggered slide-cancel. Active abilities remain a documented, conditional v1 reserve, never a silent omission.

Per-family dossiers (full morphology, biome behavior, and lineage notes) live in the Species Codex. For how these biases turn into on-the-ground rarity, see Wild Spawns.