Reading the data

Which games tell you the odds for the unit — and which only tell you the tier

Every gacha publishes a percentage. Only a handful publish one per unit. Here is exactly how precise each game's disclosure gets, and what that means for the number you are planning against.

By the GachaData team Updated 3 min read

“The 3★ rate is 3%” is a true sentence that answers a question nobody asked. What you want to know is the chance of the one unit you are pulling for, and that is a different number — usually a very much smaller one, because the tier rate is shared across everyone in the pool.

Some games publish that per-unit number. Most publish only the tier. The distinction is invisible until you put the games side by side.

How precise each game gets

GameDisclosureRows naming one unitRate rows in totalBanner rows
Blue Archiveper unit467611723
Umamusume: Pretty Derbyper unit4551,027136
Reverse: 1999per unit21021074
DIGIMON UPper unit44562
Hololive Dreamsper unit510
Arknightsby tier or banner399
CookieRun: Crumbleby tier or banner3156
Fate/Grand Orderby tier or banner210
Persona5: The Phantom Xby tier or banner62176
Azur Laneby tier or banner20
Dragon Ball Z Dokkan Battleby tier or banner15
Epic Sevenby tier or banner11
Limbus Companyby tier or banner1123
Blue Protocol: Star Resonanceby tier or banner9
Duet Night Abyssby tier or banner6
Silver And Bloodby tier or banner55
Genshin Impactby tier or banner44
Wuthering Wavesby tier or banner473
Zenless Zone Zeroby tier or banner3
Honkai: Star Railnot carried yet
Star Saviornot carried yet166

Rows naming one unit counts rows that carry a probability and identify a single member of that game’s roster. Rate rows in total includes tier bands, per-banner rates and pity tables. Banner rows is the dated banner record we keep, which is a different kind of disclosure again.

What each tier of disclosure actually buys you

Per unit — 5 games

Blue Archive, Umamusume: Pretty Derby, Reverse: 1999, DIGIMON UP, Hololive Dreams publish a probability against individual units. This is the only disclosure that answers the question directly: you can look up your target and read its chance without dividing anything by a pool size you have to count yourself.

By tier or by banner — 14 games

The majority. You are told the top-rarity rate and, usually, the featured share, and the rest is arithmetic: the chance of a specific non-featured unit is the tier rate split across however many units share that tier, which changes every time the game adds one. That is why a tier rate quietly gets worse over a game’s life without the published number ever changing.

Not carried yet — 2 games

Honkai: Star Rail and Star Savior have no rate rows here at all. That is a gap on our side, stated rather than papered over: where a number is missing we leave the cell empty instead of filling it with a plausible one. Honkai: Star Rail still has a Pull Planner, which is built from its pity model and labelled on its own page as resting on the published figures rather than on rows we carry.

Why the tier rate flatters the game

A tier rate is a promise about a category. A per-unit rate is a promise about your target. Only one of them gets smaller as the roster grows.

Take a 3% top-rarity tier shared by 110 units: an individual unit at uniform weight is 0.0273%, roughly one in 3,700 pulls. Both numbers describe the same banner. One reads as generous and the other as the actual proposition, and the game is only obliged to print the first.

Where a game publishes per-unit numbers, our pages carry them unchanged. Where it does not, the pulls-to-featured figure is the honest substitute — it folds the tier rate, the featured share and the pity together into the one quantity that is comparable across games.

What this does not say

A game low on this table is not being deceptive; per-unit disclosure is a regulatory requirement in some markets and a design choice everywhere else. The counts also measure what the game exposes for the pools it is currently running, so a game between banners legitimately shows fewer rows than one mid-event.

↑↓ to navigate · Enter to open · Esc to close