Verify LR2 HARD gauge details: guts threshold (<30 vs <32) and few-notes damage correction

[nulltask]

Context

The wave-3 LR2 alignment (commit 7bbf052) implemented the HARD gauge from beatoraja's HARD_LR2 constants. Two details lack first-party LR2 measurements and were noted as open questions in the code comments (packages/player/src/core/groove-gauge.ts):

1. Guts threshold: < 30% vs < 32%

• beatoraja master: HARD_LR2 guts {30, 0.6} (strict less-than) — currently adopted.
• lr2oraja ≥ 0.8.3 changed it to < 32 (commit message: "damage reduction start at <32%"), on the theory that LR2's 2%-granular gauge display makes "shows 30%" equal "internal < 32%".
• No first-party measurement found either way.

Action: verify against real LR2 (or a trustworthy measurement) and adjust LR2_HARD_GUTS_THRESHOLD if needed — single constant, tests pin the strict-less-than boundary.

2. Few-notes damage correction (beatoraja's fix2)

beatoraja's MODIFY_DAMAGE includes a second correction based on total note count, but executing the code as written only yields a positive factor for ~10 notes or fewer, and community analysis (ralba) describes behavior that matches a differently-bracketed loop — i.e. the implementation is suspected buggy, and whether real LR2 has a note-count correction at all is unverified. We deliberately implemented only the #TOTAL table (fix1).

Action: determine whether LR2 actually applies a note-count-based damage correction; if yes, implement it with sources.

Sources

• https://github.com/exch-bms2/beatoraja — GaugeProperty.java, GrooveGauge.java
• https://github.com/wcko87/lr2oraja (lr2oraja branch) — GaugeProperty.java, commit 3d649535
• https://ralba-gear.hateblo.jp/entry/2023/11/06/141139

[nulltask]

Part 2 (few-notes correction) resolved by first-party LR2 evidence: it exists

Reverse-engineering the actual LR2 (LR2body.exe, beta3) history.txt confirms the SURVIVAL (= our HARD) gauge scales damage by note count, and gives concrete sample points. Entry dated 080313:

SURVIVALゲージの仕様変更。
・ノート数が少ないほど、ゲージ減少量が増えます。(1000以下で順次増えて、500で20%程度、100で50%程度、20以下は即死します。)
・total値が低いほど、ゲージ減少量が増えます。

Translation: "SURVIVAL gauge spec change. (1) The fewer the notes, the larger the gauge damage — increasing progressively from ≤1000 notes: ~+20% at 500, ~+50% at 100, instant death at ≤20. (2) The lower the TOTAL, the larger the gauge damage."

So the real LR2 note-count damage curve is roughly: 1000 notes → ×1.0, 500 → ~×1.2, 100 → ~×1.5, ≤20 → death. This is in addition to the TOTAL multiplier table.

Importantly, beatoraja's fix2 does not reproduce this: executed as written it yields ~×2.0 at 500 notes, not ~×1.2. So neither our current implementation (TOTAL table only, no note-count scaling) nor a direct port of beatoraja's fix2 matches LR2. The correct curve should be derived from the LR2 sample points above.

The guts-threshold question (Part 1, <30 vs <32) is unaffected by this evidence and remains open.