Most summon-rate arguments in Anime Squadron go wrong in the same place: people reason about a single pull when the thing they actually care about is a streak. A 0.5% unit is not "one in two hundred pulls and then I have it" — it is a geometric distribution, and the median is nowhere near the mean.
The concrete version. At a 0.5% rate, the chance of seeing nothing across 200 pulls is 0.995200 ≈ 36.7%. So roughly a third of players who pull exactly "the expected amount" still walk away empty. To get to 90% confidence you need about 460 pulls; for 99% it is around 920. That gap between 200 and 920 is the entire reason banner-rate complaints feel so inconsistent between players — they are all telling the truth about their own sample.
Hard pity truncates the tail. Once a counter guarantees the unit at N pulls, the distribution stops being geometric and becomes geometric-censored-at-N, which pulls the mean down and — more importantly — collapses the variance. Soft pity (a rate that ramps as the counter climbs) does something subtler: it moves probability mass from the far tail into a narrow band just before the guarantee. Practically, that means budgeting to the soft-pity threshold is usually a better plan than budgeting to the advertised base rate.
Rates get adjusted between updates, and community-maintained numbers lag those changes. If you want to run the arithmetic against current banner values rather than against a screenshot from three patches ago, the numbers are worth recomputing — theAnime Squadron summon calculatorworks through the streak probability rather than just the average, which is the part most rate tables skip.
Published rates and datamined rates disagree often enough that it is worth knowing which one a table is built on. Published rates are per-pull; some games apply them per-slot in a multi-pull, which is not the same thing when slots are not independent. If a source does not say which convention it uses, treat its multi-pull numbers as approximate.