Deck Iteration

npm run iterate:run 5           # Run 5 iterations with auto server management
npm run iterate:status          # View current metrics
npm run iterate:generate-prompts -- --count 10  # Generate test prompts
npm run iterate:generate-decks  # Build decks from prompts
npm run iterate:analyze         # Run comprehensive analysis

Iteration Workflow

Phase 1: Preparation

1. Read ISSUES.md first: cat docs/ISSUES.md - understand known problems
2. Read iteration strategy: docs/ITERATION_STRATEGY.md
3. Check latest results: iterations/results/iteration-XXX/

Phase 2: Generate & Test

1. Choose test parameters - vary commanders, brackets (1-5), archetypes, colors, budgets
2. Generate test prompts covering diverse scenarios
3. Build decks via V4 API
4. Run validation: card count, color identity, banned cards, budget

Phase 3: Analyze Quality

Run ALL analysis tools on generated decks:

npm run dev <deck-file>           # Generate cache
npm run analyze <deck-file>       # Comprehensive analysis
npm run bracket <deck-file>       # Power level
npm run gamechangers <deck-file>  # High-impact cards

Quality dimensions:

1. Card count accuracy (exactly 100)
2. Color identity compliance
3. Format legality
4. Budget compliance
5. Bracket accuracy
6. EDHREC synergy alignment
7. Mana curve quality
8. Category distribution
9. Strategic coherence
10. Card quality/power level

Phase 4: Fix & Improve

1. Identify failures and root causes
2. Make targeted improvements to CardFiller and scoring algorithms
3. Update ISSUES.md - mark fixed issues with date, iteration, description
4. Re-test to verify improvements
5. Measure regressions in other areas

Phase 5: Record Results

• Save results to iterations/results/iteration-XXX/
• Update metrics dashboard
• Document what worked and what didn't

ISSUES.md Tracking

# When you fix an issue:

6. ✅ FIXED (Iteration 003): Too many utility lands...
    - Fixed in CardFiller by adding basic land preference scoring
    - Tested with 5 commanders, all passed

AI Multi-Analysis Optimization

For statistically optimal card selection, run reduction analysis N times with varying priorities:

1. Load mana guidelines: npm run mana-calc -- <cmc> --game-length <fast/standard/long>
2. Run N iterations (default: 10) with different strategic priorities
3. Track which cards are kept across all iterations
4. Generate consolidated "BEST" list based on frequency
5. Create timestamped output folder for all iterations

Priority variations per run:

• Synergy-focused, curve-focused, budget-focused, removal-heavy
• Combo-focused, value-focused, aggro-focused, control-focused
• EDHREC-aligned, collection-optimized

Reference

• Strategy docs: docs/ITERATION_STRATEGY.md
• Results: iterations/results/
• Known issues: docs/ISSUES.md