A result you can inspect.

Three different kinds of numbers

Wildlife measurements describe biology, with units and sources. Many legacy ABS fields still need per-field citations. A populated number is not proof of reliable evidence.

ABS ratings express editorial judgments on a 0–100 scale: attack, defense, agility, stamina, intelligence and special. They are not physical units.

Model estimates are calculated from those ratings. They do not establish what would happen in a real encounter. Read our rating definitions, source hierarchy and correction policy.

The exact preview formula

For each available factor, subtract Animal B's rating from Animal A's rating, multiply by its weight, and sum the contributions. Divide by the sum of available weights to obtain the margin.

margin = sum(weight × (A − B)) / sum(available weights)

P(A) = clamp(1 / (1 + exp(−margin / 15)), 0.05, 0.95)

Animal B receives the complementary probability. Equal ratings produce 50/50. The displayed estimate is rounded to a whole percent. The scale of 15, the weights and the 5–95% bounds are explicit design choices, not fitted scientific parameters.

Attack and defense receive the greatest weights. Agility and stamina follow; special capabilities and intelligence receive smaller weights. This is a starting hypothesis for powerscaling, not a validated account of animal combat.

No double-counting, no invented simulation count

This version uses only the six ratings. Physical mass, speed, bite force and the twelve substats are not added again as independent bonuses. Some of these already influence editorial ratings; we need a reviewed data model before adding them separately.

The result uses an analytical formula and runs zero random simulations. There is no random seed because identical inputs always produce the same output. A future sampled simulation must report its actual count and seed.

Missing measurements and confidence

Legacy physical values that are absent, invalid or zero display as Unknown. Existing zeros do not tell us whether a value was unmeasured, inapplicable or truly zero; the preview does not guess or modify those source records.

A missing rating removes that factor from both animals, and the remaining weights are renormalized. A valid rating of zero is retained. If no factors are available, the model cannot issue a probability.

Prediction confidence stays Low in this preview. Rating coverage is reported separately. More filled-in fields do not justify higher confidence without source reliability, variance, context and empirical evaluation. Extinct estimates and unsupported physical assumptions cannot establish a reliable prediction.

What is intentionally not modeled yet?

Terrain, aquatic versus terrestrial movement, ambush, starting distance, sex, prime specimens, bloodlusted behavior, groups, coordination and engagement space are not modeled. Cross-habitat results are abstract rating comparisons, not assertions that the encounter is physically plausible.

Body size is not independently scaled, so a small animal with strong relative ratings can be favored over a larger animal. This is a known limitation, not a biological conclusion. Human profiles, how-many-humans estimates and permanent editorial matchup pages require further model and evidence work.

Do not interpret community polls, tournament votes or agreement with this model as observations of real fights.

Versioning, sources and corrections

Shared results include the model and site-data release version. Unsupported old versions must be identified explicitly, never silently recomputed as historical results. Formula changes and data changes belong in the public changelog.

The model uses the repository's current animal dataset. General wildlife references on the About page are research starting points, not citations proving every field. To correct a record, provide the animal, measurement, context and exact source using the corrections process.

ABS simulations are hypothetical entertainment and powerscaling models. Never approach, provoke, fight, or test yourself against wildlife.