Sacrificial Weapon Bug: How Math.floor Broke ER

The Discovery

During a data audit of the calculation engine, we found something that didn't add up. In the code that calculates extra casts for Sacrificial weapons — one of the most widely used weapon families in Genshin Impact — the ICD cap used Math.floor. And Math.floor was producing incorrect results for most standard rotations.

The exact line in service-weapon-sacrificial.ts:

// ❌ Before:
const maxExtraFromICD = Math.floor(rotationDuration / internalCD)

// ✅ After:
const maxExtraFromICD = Math.ceil(rotationDuration / internalCD)

Why Math.floor Was Wrong

The Sacrificial weapon ICD (Internal Cooldown) starts after the first proc. The first skill hit always gets a chance to trigger the effect — it's not limited by the ICD because the ICD hasn't started yet.

In a 20s rotation with a Sacrificial Bow R1 (ICD = 30s):

• Math.floor(20 / 30) = 0 — Says: "0 extra casts max, ICD is longer than rotation"
• Math.ceil(20 / 30) = 1 — Says: "1 extra cast possible, first hit always gets its chance"

floor completely erased the first proc attempt. ceil is the correct bound: you always get at least 1 opportunity, then the ICD limits subsequent ones.

Affected Characters

This bug affects any character using Sacrificial weapons in rotations where ICD > rotation duration, or where the division isn't exact:

Character	Weapon	Refinement	ICD	Rotation	Before (floor)	After (ceil)
Xingqiu	Sacrificial Sword	R1	30s	20s	0	1
Sucrose	Sacrificial Fragments	R3	22s	20s	0	1
Diona	Sacrificial Bow	R1	30s	20s	0	1
Sayu	Sacrificial Greatsword	R1	30s	20s	0	1
Xingqiu	Sacrificial Sword	R5	16s	20s	1	2

Real Case: Sucrose in Bennett National

The most striking case was Sucrose with Sacrificial Fragments R3 in the classic Bennett National team (Xiangling / Xingqiu / Bennett / Sucrose). 20s rotation, 22s ICD:

  Sucrose (HYBRID, 80 burst, Sac Frag R3) ER:
    ❌ Before (Math.floor): 137% ER (expected range 123-151%)
    ✅ After  (Math.ceil):  117% ER (real range 105-129%)

A ~20-point ER difference. With the bug, the calculator said Sucrose needed 137% — when the real value is 117%. That's ~4 ER substats that could go into EM or Crit instead.

The other team members (Xiangling, Xingqiu, Bennett) weren't affected because their weapons aren't Sacrificial.

Why We Didn't Catch It Sooner

This code had been in place since the calculation engine rewrite. Integration tests checked ER ranges, but those ranges had been calibrated against the incorrect engine behavior. It took a manual code audit — comparing the ICD logic against actual in-game mechanics — to spot the discrepancy.

Once we identified the bug, 7 tests failed. All were Sucrose. All expected the wrong value. We updated the fixtures and now all 894 tests pass with the correct behavior.

How to Verify Your Builds

If you use ER QuickCalc with Sacrificial weapon characters, your calculations already use the corrected value. Just re-calculate your team and the new numbers reflect the fix.

To verify manually:

1. Select a character with a Sacrificial weapon (Xingqiu, Sucrose, Diona, Sayu)
2. Set up your team in ER QuickCalc
3. Check the recommended ER — if it dropped vs your previous build, the fix is working correctly