The unlucky tail: the number the average hides
Two games can average the same number of pulls and cost wildly different amounts when it goes badly. The gap between a typical run and an unlucky one is the number worth planning against, and here it is for every game.
Every gacha comparison you will read leads with an average. It is the wrong number to save against.
An average is what happens across thousands of accounts. What happens to yours is one draw from a distribution, and in a gacha that distribution is badly lopsided: a long right tail where the unlucky runs live. Plan against the average and you will be short roughly half the time.
The useful pair of numbers is the typical run and the unlucky one — the 50th and 90th percentile. One in ten players does worse than the unlucky column.
Typical, average, unlucky
| Game | Typical (50%) | Average | Unlucky (90%) | Guaranteed by | Spread |
|---|---|---|---|---|---|
| CookieRun: Crumble | 116 | 167 | 383 | never | 267 |
| Dragon Ball Z Dokkan Battle | 97 | 140 | 322 | never | 225 |
| Umamusume: Pretty Derby | 93 | 133 | 306 | not carried | 213 |
| Fate/Grand Order | 87 | 116 | 287 | 330 | 200 |
| Azur Lane | 58 | 76 | 191 | 200 | 133 |
| Hololive Dreams | 69 | 87 | 200 | 200 | 131 |
| Limbus Company | 96 | 106 | 200 | 200 | 104 |
| Blue Archive | 99 | 108 | 200 | 200 | 101 |
| Arknights | 57 | 69 | 142 | never | 85 |
| Genshin Impact | 80 | 91 | 156 | 180 | 76 |
| Silver And Blood | 36 | 50 | 111 | never | 75 |
| Duet Night Abyss | 81 | 94 | 156 | 180 | 75 |
| Honkai: Star Rail | 81 | 94 | 156 | 180 | 75 |
| Zenless Zone Zero | 81 | 94 | 156 | 180 | 75 |
| Persona5: The Phantom X | 74 | 88 | 145 | 160 | 71 |
| Wuthering Waves | 75 | 83 | 145 | 160 | 70 |
| DIGIMON UP | 139 | 127 | 200 | 200 | 61 |
| Reverse: 1999 | 70 | 65 | 123 | 140 | 53 |
| Blue Protocol: Star Resonance | 68 | 65 | 71 | 80 | 3 |
All figures are pulls to the banner’s featured unit from zero, with no carried-over pity, and already account for each game’s ramp or spark. Spread is the unlucky figure minus the typical one — how much extra a bad run costs. Sorted widest spread first.
The spread is the story
A guarantee is a spread cap
Look at what the Guaranteed by column does to the Spread column. The games with no guarantee occupy almost the whole top of this table. Where a game guarantees the unit, the unlucky case cannot run past the ceiling, so the spread is bounded by construction. Where it does not, nothing bounds it — the 90th percentile is simply where the distribution happens to be, and the 99th is further still.
This is why a guarantee is worth more than a good average: it converts an open-ended risk into a known worst case you can actually save for.
Two games can tie on the average and not be alike
The averages compress games that behave very differently. Read the typical and unlucky columns together and pairs that look identical in a one-number ranking come apart immediately — a game whose ramp is tight lands nearly everyone in a narrow band, while a flat-rate game scatters the same average across a range several times wider.
What to actually budget
If you are saving for a specific unit, the unlucky column is the honest target and the guarantee column is the ceiling. The average is useful for comparing games, and for nothing else. Each game’s own Pull Planner runs these numbers for any starting pity count you already have.
What this does not say
These are pull counts, not money: a pull costs a different amount in every game, which is its own comparison. The percentiles assume you start from zero pity and stop the moment you get the unit. And they describe the banner type each game leads with — a game running several pools with different rules is represented by one of them, named on its own page.