How we verify data

Sourcemined reads data directly from the files each game ships with, on a copy we bought. Nothing on the site is typed in by hand or copied from another wiki.

Extraction

For Mewgenics, our importer opens resources.gpak, parses the game's own data files (GON and CSV), and joins them with the official translations. Every page links back to the file and entry it came from.

Patch stamps

Each page shows the game build it was extracted from. When a patch ships, we re-run the importer, publish a list of what changed, and update every affected page.

Mechanics we reconstruct

Some rules, like breeding inheritance, aren't in the data files. When we describe those, we explain them in our own words and math, state our confidence, and never publish the game's code.

Breeding

These rules power the breeding planner. We derived them from the game's files for build 25143593 and re-verify them on every patch. Each rule is marked verified (read directly from the game), inferred (reconstructed and consistent with play, but not confirmed) orunknown.

Stats verified

A kitten never averages its parents' stats. Each of the seven stats is copied whole from one parent. Both parents start with equal weight, so every stat is a 50/50 pick.

Each point of the room's Stimulation adds 0.01 to the weight of whichever parent has the higher value for that stat. At Stimulation 50, the better parent's value is picked 60% of the time; it never reaches 100%.
Negative Stimulation favors the worse parent the same way.
The chance of taking a parent's value is that parent's weight divided by the sum of both weights.
The game also supports favor mom, favor dad and favor best furniture effects (0.25, 0.25 and 0.1 per point), but no furniture currently has them.

Abilities and passives verified

First ability: 20% plus 2.5% per point of Stimulation, so it's certain from Stimulation 32.
Second ability: 2% plus 0.5% per point of Stimulation.
Passive: 5% plus 1% per point of Stimulation.

Which parent the ability or passive comes from is inferred: a weighted pick like stats, where favor class effects push it toward a parent that has a class ability or passive.

Disorders and birth defects verified

Each parent's disorder passes on with a 15% chance. Stimulation doesn't change this.
A new birth-defect disorder appears with 2% plus 40% of however much the parents' kinship exceeds 0.2. It is not rolled if the kitten already inherited two disorders.
When kinship is 0.05 or more, a body part becomes a birth-defect mutation with a chance of 1.5 × kinship. Above a kinship of 0.9, two parts are affected; that part count is inferred.

Kinship verified

Relatedness is the standard coefficient of kinship over the family tree: the chance that a gene picked at random from each parent is the same inherited copy. Siblings and parent-child pairs are 0.25, half-siblings 0.125 and first cousins 0.0625. Inbred lines push it higher.

Litters verified

Every cat has a hidden fertility value between 1.0 and 1.25: the lower of two random rolls, so about 1.08 on average. A kitten is conceived with a chance equal to the two cats' fertility multiplied together, so with any two cats the first kitten is certain. The amount above 1 is the chance of a twin, at most 56% for two cats at 1.25. Same-sex pairs never conceive unless one of the cats is neutral. We checked the fertility range against three players' saves.

Who breeds with whom verified

Each night, cats in the same room meet and decide whether to breed. How keen one cat is on another is 0.15 × the other cat's Charisma × its own libido × an orientation factor × a lover factor:

Orientation depends on the cat's hidden sexuality, from 0 (straight) to 1 (gay). Interest in the opposite sex falls as sexuality rises and interest in the same sex grows, smoothly along a quarter circle. A neutral cat on either side removes the factor.
Lovers: a cat with a lover is more interested in that lover (1 + how attached it is) and less in everyone else (1 − that attachment).
Both cats roll, one after the other, and both must agree. Each roll succeeds with that keenness × √(1 + 0.1 × Comfort), so a comfortable room makes breeding more likely and a room at −10 Comfort or lower stops it.
A room with a Breed Suppression item, such as the Idol of Chastity, doesn't breed at all.

Charisma here is the cat's total, including bonuses. A loaded save gives us level-ups and injuries but not equipment or mutation bonuses, so the planner's Charisma, and the "agree" chance, can be slightly off for geared cats. Which cats meet each night isn't modelled: the planner shows how likely a pair is to breed if they meet.

Rooms and furniture verified

A room's stats are the sum of its furniture's effects. The breeding rules read the same Stimulation number the game shows for the room.
Comfort drops by 1 for each cat over 4 in the room.
A rare piece of furniture has double Comfort, Appeal, Health, Stimulation and Evolution. Its other effects, like food storage, are unchanged.
The game supports furniture that favors a parent, raises ability odds or prevents birth defects, but no furniture in the current build has any of those effects. That's why the planner only asks for Stimulation and Comfort.

Reading your save

The planner reads your save file in the browser: it is never uploaded. We decode the parts the planner needs (cats, the family tree, the house and its furniture). We checked the decoding against three players' saves: the inbreeding we compute matches the value the game stored for every one of 2,135 kittens, and kittens' stats and body parts come from their parents as the rules predict.

Found a mistake?

Tell us on Discord with the page link and the value you see in game, and we'll check it against the files.

Hytale

For Hytale, our importer reads Assets.zip from the game install, resolves every asset's parent chain the way the game server does, and joins the result with the seven official translations. Each page shows the game version it came from.

Inherited fields

Many items are variants of a template: they start from the template's values and override some of them. Two fields are never inherited, which some guides get wrong: an item's recipe and its quality. A variant without its own recipe can't be crafted, even if its template can. Attack and interaction lists combine entry by entry, so a variant can change one attack and keep the rest.

Drop chances verified

Drop chances are calculated exactly from the drop tables, not estimated from kills. A drop table is a tree of four kinds of entries:

Single: drops one item. The amount is uniform between the minimum and maximum, so 1–3 means one, two or three with equal odds.
Multiple: every entry inside rolls on its own, with a chance of its weight divided by 100 (a weight of 100 always drops). The group can repeat; the repeat count is a random point between its minimum and maximum rounded to the nearest whole number, which makes the two end values half as likely as the ones between. For 1–3 repeats that's 25% one, 50% two and 25% three.
Choice: picks exactly one entry per roll, with odds proportional to the weights. The number of rolls is uniform between its minimum and maximum.
Linked table: rolls another drop table. A table only fires once per kill or break, even if it's linked twice. None of the tables in the current version links the same table twice, so every chance on the site is exact.

For each item we show the chance of getting at least one per kill or break, and the average amount. The average adds up all the ways the item can drop, so it can exceed 1 even when the chance is below 100%.

Crafting calculator

The calculator uses each item's own recipe when it has one, otherwise another recipe that produces it, like a furnace. It adds up shared ingredients before rounding to whole crafts, never uses salvaging or unfinished recipes, and treats "any item of a family" inputs and loops as raw materials.

Graveyard Keeper 2

For Graveyard Keeper 2, our importer reads the game's balance data from the install and joins it with the official translations. Each page shows the build it came from.

Star quality verified

A star craft has a length in ticks and a mastery requirement in one of the mastery colours. Every hit of the worker moves it forward:

At or above the requirement, every tick succeeds. A hit does one tick, two at double the requirement and three at triple (never more), so high mastery mostly makes a craft faster.
Below the requirement, a hit does one tick, and it succeeds with a chance close to your mastery divided by the requirement. The game rolls a whole number from 1 to 99, so the exact chance is the share of those numbers at or under 100 × mastery ÷ requirement. A failed tick still counts towards finishing.
Head start: every ingredient above one star adds a tick that's already a success (a two-star ingredient one, a three-star ingredient two), and some perks linked to the craft add more. Others make the craft longer, which gives more chances to succeed.
Zombies always count as meeting the requirement, so a zombie's result never varies.

The result depends only on the number of successful ticks: enough for the craft's bronze, silver or gold level gives one, two or three stars. Each tick is independent, so we add up every possible number of successes exactly (no sampling) and check the formula against a hit-by-hit simulation of the same rules.

Perk-dependent amounts inferred

Many recipes give a different amount, or take a different time, depending on a perk: the data stores these as small formulas like "four, plus two with the cooking perk". We read each perk in a formula as one when you have it and zero when you don't. That gives sensible amounts for every recipe, but we haven't confirmed it in game yet, so treat amounts that depend on a perk with a little care.

Tech paths inferred

We treat a tech with more than one parent in the tree as needing all of them. The cost of reaching a tech is the sum of the red, green and blue points of the tech and every tech above it, each counted once. Thetech planner leaves out what you already have and everything it needed.

Crafting planner

The planner uses, for each item, the first craft that makes it for certain. Studies, autopsies and outputs that only drop by chance are never used, inputs that accept any item of a group stay as raw materials, and shared ingredients are added up before rounding to whole crafts.