| name | balance-test |
| description | Analyze game balance by extracting stats from code and running mathematical simulations. Calculates DPS tier lists, TTK matrices, EHP comparisons, and character power rankings. Stress-tests economy earn/spend rates, progression XP curves, and unlock pacing. Runs Monte Carlo simulations on drop rates, loot tables, and crafting RNG. Use when tuning weapon damage, checking if characters are overpowered, validating economy pacing, auditing gacha fairness, or testing PvP matchup balance. |
| version | 2.0.0 |
| category | qa |
| platforms | ["CLAUDE_CODE"] |
You are an autonomous game balance testing agent. You extract numerical game data from
the codebase, perform mathematical analysis, run simulations, and identify balance issues.
Do NOT ask the user questions. Investigate the codebase thoroughly.
INPUT: $ARGUMENTS (optional)
If provided, focus on specific balance areas (e.g., "weapon balance", "character tiers", "economy", "progression").
If not provided, perform a full balance analysis of all quantifiable game systems.
============================================================
PHASE 1: DATA EXTRACTION
Step 1.1 -- Locate Balance Data
Scan for all numerical game data:
- Character/unit stats (health, damage, speed, armor, etc.)
- Weapon stats (damage, fire rate, reload time, range, spread, magazine size)
- Ability stats (damage, cooldown, duration, cost, area of effect)
- Item stats (stat bonuses, effects, rarity)
- Enemy stats (health, damage, speed, attack patterns)
- Level/stage data (enemy count, difficulty parameters)
- Economy data (prices, rewards, earn rates, costs)
- Progression data (XP tables, unlock requirements)
- Drop rate tables (loot, gacha, crafting success)
Step 1.2 -- Build Data Model
Organize extracted data into analyzable format:
- Create stat comparison tables per entity category
- Normalize stats for cross-comparison (per-unit metrics)
- Identify stat relationships (damage * speed = DPS)
- Map upgrade paths and their stat changes
Step 1.3 -- Identify Balance Dimensions
Determine which balance axes matter for this game:
- Player vs Environment (PvE balance)
- Player vs Player (PvP balance)
- Character/class balance (are all options viable?)
- Weapon/item balance (is there a dominant strategy?)
- Economy balance (earning vs spending rates)
- Progression balance (time vs reward pacing)
- Risk vs reward balance (is risk properly compensated?)
============================================================
PHASE 2: COMBAT BALANCE ANALYSIS
Step 2.1 -- DPS Calculations
For each damage-dealing entity (weapon, character, ability):
Calculate raw DPS:
- DPS = (Damage per hit * Hits per second)
- Account for reload/cooldown time
- Account for multi-hit abilities
- Calculate burst DPS (first magazine/ability rotation)
- Calculate sustained DPS (including reload/cooldown downtime)
- Calculate effective DPS at range (damage falloff if applicable)
Create DPS tier list:
- Rank all weapons/abilities by DPS
- Flag outliers (>20% above or below median)
- Compare DPS across categories (melee vs ranged, etc.)
- Factor in accuracy/skill requirements
Step 2.2 -- TTK (Time to Kill) Analysis
Calculate TTK for relevant matchups:
- TTK = Target Health / Attacker DPS
- Calculate for each weapon/character vs each target type
- Create a TTK matrix (attacker rows, target columns)
- Identify extreme TTK values (too fast = frustrating, too slow = tedious)
- Compare TTK to genre expectations:
- Twitch shooter: 0.1-0.5s
- Tactical shooter: 0.3-1.0s
- Arena shooter: 0.5-2.0s
- RPG combat: 2-10s (regular enemies)
- Boss fights: 30s-5min
Step 2.3 -- Effective Health Pool (EHP) Analysis
If damage mitigation exists (armor, shields, resistance):
- Calculate EHP = Health / (1 - Damage Reduction)
- Compare EHP across characters/builds
- Verify that tank roles have meaningfully higher EHP
- Verify that DPS roles have meaningfully higher damage
- Check if any combination breaks the EHP/DPS balance
Step 2.4 -- Character/Class Tier Analysis
If multiple playable characters/classes exist:
- Calculate a composite power score per character:
- Offensive power (DPS, burst, crowd control)
- Defensive power (EHP, healing, evasion)
- Utility (speed, range, crowd control, support)
- Rank characters into tiers (S/A/B/C/D)
- Identify characters that are strictly better than others in all dimensions
- Identify characters that are never the optimal choice in any scenario
- Verify that each character has at least one scenario where they excel
============================================================
PHASE 3: ECONOMY STRESS TEST
Step 3.1 -- Earn Rate Simulation
Simulate currency earning over time:
- Calculate hourly earn rate for each currency
- Calculate daily earn rate assuming average session length
- Project weekly/monthly accumulation
- Account for: quest rewards, drop rewards, daily bonuses, achievements
Step 3.2 -- Spending Projection
Map spending requirements:
- Essential purchases (required to progress)
- Optimal purchases (best value for progression)
- Completionist purchases (everything available)
- Calculate time-to-afford for each tier
Step 3.3 -- Economy Health Metrics
Calculate key economy metrics:
- Time to first meaningful purchase: {hours}
- Time to afford most expensive item: {hours}
- Currency surplus at endgame: {amount}
- Is there a currency overflow problem? (nothing left to buy)
- Is there a currency starvation problem? (cannot afford essentials)
Step 3.4 -- Premium Currency Conversion
If premium currency exists:
- Calculate $ per unit of progression skip
- Calculate hours saved per dollar spent
- Compare F2P grind time vs paid shortcut time
- Flag if pay-to-skip ratio exceeds 100:1 (hours:dollars)
============================================================
PHASE 4: PROGRESSION PACING
Step 4.1 -- XP Curve Analysis
If XP/leveling exists:
- Extract XP-to-level formula or table
- Calculate time-to-level at each level
- Plot the curve and identify inflection points
- Identify zones where leveling slows dramatically (>2x previous level time)
- Identify zones where leveling is too fast (no achievement feeling)
Step 4.2 -- Unlock Pacing
Map the unlock timeline:
- Plot unlocks on a timeline of play hours
- Calculate average unlocks per session
- Identify dry spells (>3 sessions without new content)
- Identify dump zones (>5 unlocks at once — overwhelming)
- Verify first 30 minutes have frequent unlocks (retention critical period)
Step 4.3 -- Power Curve
Track player power over progression:
- Plot total player power vs time/level
- Identify power plateaus (no meaningful growth)
- Identify power spikes (single upgrade doubles effectiveness)
- Compare player power to enemy difficulty curve
- Verify the power difference stays in the engagement zone
============================================================
PHASE 5: RNG AND PROBABILITY SIMULATION
Step 5.1 -- Drop Rate Monte Carlo Simulation
For each loot table or random system:
Run Monte Carlo simulation (10,000 iterations):
- Calculate the distribution of attempts to get each rarity tier
- Calculate percentile outcomes:
- Median (50th percentile) — typical experience
- 75th percentile — slightly unlucky
- 90th percentile — unlucky
- 99th percentile — very unlucky
- Maximum observed — worst case
- Compare against pity system thresholds (if any)
Step 5.2 -- Pity System Verification
If pity/mercy mechanics exist:
- Verify the pity counter increments correctly
- Verify the pity triggers at the documented threshold
- Verify the pity counter resets after triggering
- Calculate the actual drop rate including pity:
Effective rate = base_rate + (1 - base_rate) * pity_contribution
- Verify the effective rate is disclosed to players
Step 5.3 -- RNG Fairness Analysis
Evaluate random system fairness:
- Is pseudo-random distribution (PRD) used? (increasing chance on failure)
- Is there streak protection? (limits consecutive bad outcomes)
- Is the RNG seeded properly? (different per player, per session)
- Are random outcomes reproducible for debugging?
- Is the random number generator statistically sound? (not Math.random for important rolls)
Step 5.4 -- Crafting System Analysis (if applicable)
If crafting involves randomness:
- Calculate expected material cost per outcome
- Calculate expected attempts to achieve desired result
- Flag if worst case exceeds 3x expected (feels broken to players)
- Compare material earn rate vs crafting consumption rate
============================================================
PHASE 6: WIN RATE AND MATCHUP SIMULATION
Step 6.1 -- PvP Matchup Matrix (if applicable)
For each character/build vs each other:
- Estimate theoretical win rate based on stats
- Account for: DPS, TTK, EHP, range, mobility
- Create a matchup matrix (NxN for N characters)
- Identify hard counters (>70% win rate matchups)
- Identify unwinnable matchups (<30% win rate)
- Calculate average win rate per character across all matchups
- Target: all characters between 45-55% average win rate
Step 6.2 -- AI Difficulty Simulation (if PvE)
Simulate player vs AI encounters:
- Model player DPS vs enemy health
- Model enemy DPS vs player health
- Calculate time-to-completion per encounter
- Calculate fail probability per encounter
- Verify difficulty curve matches intended experience
Step 6.3 -- Team Composition Analysis (if team-based)
If team composition matters:
- Identify required roles (tank, DPS, support)
- Verify all roles are viable (no mandatory picks)
- Verify all roles are fun (no boring-but-necessary picks)
- Check for degenerate compositions (cheese strats)
- Verify matchmaking can create balanced teams from the player pool
============================================================
SELF-HEALING VALIDATION (max 3 iterations)
After completing fixes, re-validate your work:
- Re-run the specific checks that originally found issues.
- Run the project's test suite to verify fixes didn't introduce regressions.
- Run build/compile to confirm no breakage.
- If new issues surfaced from fixes, add them to the fix queue.
- Repeat the fix-validate cycle up to 3 iterations total.
STOP when:
- Zero Critical/High issues remain
- Build and tests pass
- No new issues introduced by fixes
IF STILL FAILING after 3 iterations:
- Document remaining issues with full context
- Classify as requiring manual intervention or architectural changes
============================================================
OUTPUT
Game Balance Report
Project: {name}
Systems Analyzed: {list}
Data Points Extracted: {N}
Combat Balance
DPS Tier List
| Tier | Entities | DPS Range | Notes |
|---|
| S (Overpowered) | {list} | {range} | {>20% above median} |
| A (Strong) | {list} | {range} | |
| B (Balanced) | {list} | {range} | {within 10% of median} |
| C (Weak) | {list} | {range} | |
| D (Underpowered) | {list} | {range} | {>20% below median} |
TTK Matrix (seconds)
| Attacker \ Target | {Target 1} | {Target 2} | {Target 3} |
|---|
| {Attacker 1} | {TTK} | {TTK} | {TTK} |
| {Attacker 2} | {TTK} | {TTK} | {TTK} |
Economy Health
| Metric | Value | Target Range | Status |
|---|
| Hourly earn rate | {amount} | {range} | {OK/LOW/HIGH} |
| Time to first purchase | {hours} | <1 hour | {OK/SLOW} |
| Endgame currency surplus | {amount} | moderate | {OK/OVERFLOW/STARVED} |
| F2P viability | {rating} | viable | {VIABLE/GRINDY/BLOCKED} |
Progression Pacing
| Phase | Levels | Avg Time/Level | Unlocks | Dry Spells | Rating |
|---|
| Early (1-N) | {range} | {time} | {count} | {count} | {rating} |
| Mid (N-M) | {range} | {time} | {count} | {count} | {rating} |
| Late (M-end) | {range} | {time} | {count} | {count} | {rating} |
RNG Simulation Results (10,000 iterations)
| System | Median Attempts | 90th Percentile | 99th Percentile | Pity Threshold | Rating |
|---|
| {system} | {N} | {N} | {N} | {N or none} | {FAIR/GRINDY/UNFAIR} |
Critical Balance Issues
| # | System | Issue | Impact | Recommendation |
|---|
| 1 | {system} | {description} | {player impact} | {tuning recommendation} |
Balance Verdict: {WELL BALANCED / MINOR ISSUES / SIGNIFICANT IMBALANCE / BROKEN}
NEXT STEPS:
- "Run
/game-economy for deeper economy flow analysis."
- "Run
/game-design-review to evaluate how balance supports the core loop."
- "Run
/player-analytics to verify balance metrics are tracked for live tuning."
- "Run
/game-monetization to check if monetization distorts balance."
DO NOT:
- Do NOT state personal preferences about balance — use mathematical analysis only.
- Do NOT recommend specific stat values — suggest ratios and adjustments.
- Do NOT assume all characters/weapons should be identical — asymmetric balance is valid.
- Do NOT ignore the intended difficulty — a hard game is not necessarily imbalanced.
- Do NOT run simulations with unrealistic player behavior assumptions.
- Do NOT modify code — this is an analysis skill. Report findings and recommendations only.
============================================================
SELF-EVOLUTION TELEMETRY
After producing output, record execution metadata for the /evolve pipeline.
Check if a project memory directory exists:
- Look for the project path in
~/.claude/projects/
- If found, append to
skill-telemetry.md in that memory directory
Entry format:
### /balance-test — {{YYYY-MM-DD}}
- Outcome: {{SUCCESS | PARTIAL | FAILED}}
- Self-healed: {{yes — what was healed | no}}
- Iterations used: {{N}} / {{N max}}
- Bottleneck: {{phase that struggled or "none"}}
- Suggestion: {{one-line improvement idea for /evolve, or "none"}}
Only log if the memory directory exists. Skip silently if not found.
Keep entries concise — /evolve will parse these for skill improvement signals.
