Methodology
The exact pity model behind HOYO WISH: base rates, soft pity ramp, hard pity, 50/50 guarantee — plus sources and the [VERIFY] policy.
The shared pity model
rate(n) = base + boost × max(0, n − softPity) capped at 100% P(within k) = 1 − (1 − r₁)(1 − r₂)…(1 − rₖ) expected pulls = Σ P(still nothing after i−1)
Default parameters for the character-banner preset [VERIFY per game & patch]: base rate 0.6%, soft pity starts ~pull 74, boost +6% per pull (community estimate), hard pity 90. Weapon/light-cone style banners use 1.6% base and 80 hard pity; the calculator exposes every parameter.
The 50/50
With guarantee off, every 5★ is the rate-up with probability 0.5; a loss guarantees the next. Long-run rate-up odds sit near 55% under this model [VERIFY]. The planner treats a carried guarantee as probability 1.
Currency conversions
Primogem-style conversions (e.g. 160 primogems per pull) are editable placeholders [VERIFY per game].
The [VERIFY] policy
Every game value carries a flag and the dataset a review date (currently 2026-10-08). Patch-day flow: re-check sources → update src/data/site.ts → bump date → changelog.
AI-assist disclosure
Pages drafted with AI assistance, reviewed by the operator before publishing. All math is deterministic client-side code.