Methods, sources and limits
How Palbreedery Verifies Its Breeding Data
A green check is only useful when the reference is independent. This page records where the shipping table comes from, what is compared, the failure that changed our method, and what the calculator still cannot prove.
Breedable roster
287
Shipped table cells
41,328
Independent checks
82,367
Three separate jobs, three named sources
Breeding results
The table shipped to the browser is derived from a Palworld 1.0 game-file extraction made with CUE4Parse.
tylercamp/palcalc ↗Ranks and names
Breed ranks and English display names are selected facts from DT_PalMonsterParameter and the game localization data.
palworld-save-pal ↗Independent check
At verification time, the shipping engine is compared with a separate project's published matrix. Its data is fetched for the check and is not redistributed.
Kimpton calculator ↗Licence and attribution details are preserved in the project's third-party notices.
What the verification actually checks
Normal directional results
82,367 applicable cells are compared against the independent matrix. The two directions are checked even though the shipped table stores an unordered pair once.
Gender-dependent outcomes
Female Katress + male Wixen and the reversed genders are checked separately because they produce different children.
Named regression case
Chikipi + Cremis must return Foxparks. This pins the tie case that the earlier formula implementation got wrong.
Table structure
The verifier checks the upper-triangle size, the single gender-dependent hole, index bounds, aliases and the excluded invalid entry.
The failure that changed the method
The first engine recalculated offspring from breed ranks and disagreed with the later independent matrix on 47,581 of 82,944 directional results (57.4%). A reversed equal-distance tie was one demonstrated failure, but it was not the only source of disagreement. Chikipi (3080) + Cremis (2890) targets 2985: Fuack at 2980 and Foxparks at 2990 are equally close; the old engine returned Fuack while the extracted table returns Foxparks.
Worse, the old verification compared the engine with the same upstream used to build its overrides. It reported a perfect pass while the answers were wrong. Palbreedery now ships the extracted result table instead of re-deriving the game rule, and verifies that shipping code against a separate extraction.
Why the calculator lists 287 breedable Pals
A Paldeck count is not automatically a breeding roster. The shipped set excludes tower bosses marked with the non-breedable rank, a duplicate cosmetic Gumoss entry, Terraria crossover monsters that form a closed non-Pal group, and Astralym, which is flagged invalid or unavailable upstream. Excluding those entries does not remove any ordinary-Pal pair result from the table.
Known limits
No in-game hatch test
Two independent breeding-result extractions agree across the shipped roster, but a reproducible in-game contradiction should override them and be reported.
No inheritance odds
Mutation and genetic recombination probabilities are omitted because this dataset does not contain them and Palbreedery has not independently validated a figure.
No Palbox route planner
The site returns direct pairs. It does not yet calculate the shortest multi-generation chain from the Pals a player owns.
Dataset version: 1.0. After a game update, the correct response is to regenerate and rerun verification, not to assume the previous table still applies.