9 formulas, 7 hypotheses, 1 conflict: the damage spec lands with caveats
docs/mechanics/damage-and-success.md is complete: 440 lines, 9 formulas,
18 rules, 7 HYPOTHESIS and 2 UNVERIFIED formulas out of 9 total.
Engine baseline: 1 test passed in 0.01s unchanged.
The tiomasta problem
tiomasta’s GameFAQs thread [S-B] is the only source publishing actual damage formulas for Medabots GBA. It claims: Evasion and Defense each subtract 25% of their value from damage (F2), ROS adds 25% of itself as bonus damage (F1), and crits are deterministic ROS vs (Evasion + Defense) with a small random factor (F4). Methodology: unstated. Every formula from this source is therefore HYPOTHESIS-grade until disassembly or in-game measurement confirms it.
This is the core tension in T0.4: we need formulas to build the engine, but the only formulas available are community-tested without documented methods. The spec therefore explicitly labels each formula’s provenance and provides falsifiable test cases for future verification.
What’s verified vs. hypothesized
DOCUMENTED (4 rules):
- Armor absorption and part destruction (rules 9-11) [S-A, battle-flow T0.2]
- Leader/partner head destruction outcomes (rules 17-18) [battle-flow T0.2]
HYPOTHESIS (11 rules, 7 formulas):
- F1: BaseDamage = Power + (Power × SkillLevel/100) + (Success × 0.25)
- F2: DamageReduction = (Evasion × 0.25) + (Defense × 0.25); FinalDamage = max(1, BaseDamage - DamageReduction)
- F3: Hit if Success >= Evasion
- F4: Crit if Success >= Evasion + Defense
- F6: Excess damage passes through to another part on destruction
- F7: Recovery-turn (RAD phase) attack-type weaknesses exist
- F9: Anti-Air vs Flying and Anti-Sea vs Aquatic deal 2× damage
UNVERIFIED (3 rules):
- Crit damage multiplier (assumed 2×, community convention)
- Passthrough priority (which part absorbs excess damage)
- Recovery-turn weakness specifics (which types, multipliers)
The compatibility conflict
Medapedia documents that Medal-part Attribute compatibility exists [S-A]. Kenshin_Xtreme’s guide states Tortoise/Jellyfish/Bear medals give +1 stat bonus. The owner-supplied compiled report claims +7 for the same medals. This is a direct conflict recorded in BACKLOG.md [2026-09-23]. Until disassembly or in-game measurement resolves this, do not implement the magnitude in the engine. The existence of compatibility is DOCUMENTED; the bonus value is UNVERIFIED.
Numbers from this session
- 440 lines of spec in
docs/mechanics/damage-and-success.md— the longest spec so far, surpassing T0.3’s 419 lines by 21 lines. - 9 formulas documented across F1-F9, with 7 HYPOTHESIS and 2 UNVERIFIED (crit multiplier, passthrough priority).
- 18 rules in numbered pseudocode form for engine implementation.
- 6 derived test oracles for Phase 1 engine invariants.
- 5 sources cited (S-A through S-E), with S-B being the primary formula source.
Falsifiable predictions
Each HYPOTHESIS formula includes a “Test to verify” section. Examples:
- F2: Attack with BaseDamage=100 against Evasion=100, Defense=0 should deal 75 damage (100 - 25). Same against Evasion=0, Defense=100: 75 damage. Against Evasion=100, Defense=100: 50 damage.
- F4: Part with Success=100 always crits targets with Evasion+Defense ≤ 100. Success=100 vs Evasion=40, Defense=50 (threshold=90): critical hit.
- F1: Head with Power=100, Success=40, Medal Skill Level=100 (Shoot): BaseDamage = 100 + (100×1) + (40×0.25) = 210.
Session artifacts
- Spec file:
docs/mechanics/damage-and-success.md(SHA256:0547dbb51f85309127165cb357aec31e9715d2f6743ef3a27667e1600227c790) - Session report:
temp/session_report.jsonwith matching metrics - Engine test: 1 baseline test passed in 0.01s
What’s next
T0.4 acceptance criteria are met: every formula has provenance and a proposed in-game test. The spec is ready for T0.7 data schema work (which needs these formula stat fields) and T1.3 turn loop implementation (which needs these formulas for damage resolution).
The biggest gap is disassembly verification of tiomasta’s claims. Reading medarot1’s battle routines (S-C) would elevate HYPOTHESIS formulas to VERIFIED or reveal corrections. This is now the highest-priority research task for Phase 0 completion.
Claim labels: DOCUMENTED claims cite S-A or battle-flow T0.2. HYPOTHESIS claims cite S-B. UNVERIFIED claims have conflicting or missing sources.