# Archived Documentation

**Historical documentation for deprecated features and old system versions.**

This directory contains documentation for features that have been removed, deprecated, or replaced by newer systems. Kept for historical reference and learning from past decisions.

---

## 📦 What's Archived Here

**Archived files are kept for:**
1. **Historical reference** - Understanding past design decisions
2. **Learning from failures** - Why certain approaches didn't work
3. **Baseline comparisons** - Comparing new systems to old
4. **Recovery information** - If rollback ever needed (rare)

**Archived files are NOT:**
- ❌ Active in production
- ❌ Maintained or updated
- ❌ Recommended for new implementations
- ❌ Complete or fully accurate anymore

---

## 🗄️ Currently Archived (Examples)

### **Indicator Versions (Replaced by v9)**

**v5: Buy/Sell Signal Strategy**
- **Period:** Pre-Nov 12, 2024
- **Performance:** 36.4% WR, +$25.47 total
- **Why replaced:** Poor win rate, basic signal detection
- **Successor:** v6 HalfTrend + BarColor

**v6: HalfTrend + BarColor**
- **Period:** Nov 12-18, 2024
- **Performance:** 48% WR, -$47.70 total
- **Why replaced:** Marginal improvement, still underperforming
- **Successor:** v8 Money Line Sticky Trend

**v7: v6 with Toggles**
- **Period:** Nov 2024 (minimal data)
- **Performance:** Deprecated before significant data collected
- **Why replaced:** No fundamental improvements over v6
- **Successor:** v8 Money Line Sticky Trend

**v8: Money Line Sticky Trend**
- **Period:** Nov 18-26, 2024
- **Performance:** 57.1% WR, +$262.70 (8 trades)
- **Why replaced:** SHORT failures (5 oversold disasters), weak chop entry
- **Key issues:**
  * RSI filter for SHORTs caught oversold (25-35 RSI) = wrong entries
  * Quality 80 signal at ADX 20.7 (weak chop) = stopped out
- **Successor:** v9 Money Line with Momentum-Based SHORT Filter
- **Lesson learned:** Oversold doesn't mean reversal, need momentum at tops

**v10: Adaptive Position Scoring (REMOVED)**
- **Period:** Nov 28, 2024 (1 day)
- **Performance:** Parameter insensitivity (72 configs = identical $498.12 P&L)
- **Why removed:** Bug in penalty logic, no edge over v9, added complexity
- **Key issues:**
  * Price position penalty applied to 18.9% (should only apply to 40-60% chop)
  * No performance improvement despite added complexity
  * Violated "simpler is better" principle
- **Lesson learned:** Parameter insensitivity indicates no real edge, just noise
- **Status:** ✅ FULLY REMOVED (Nov 28, 2025)

---

## 🎯 Current Production System

**v9: Money Line with MA Gap + Momentum SHORT Filter**
- **Period:** Nov 26, 2025+
- **Status:** ✅ PRODUCTION
- **Key Features:**
  * MA Gap analysis (+5 to +15 quality points)
  * Momentum-Based SHORT filter (ADX ≥23, Position ≥60% OR ≤40% + Vol ≥2.0x)
  * Removed RSI filter for SHORTs (data showed RSI 50+ has 68.2% WR)
  * Quality thresholds: LONG ≥90, SHORT ≥80
- **File:** `workflows/trading/moneyline_v9_ma_gap.pinescript`
- **Documentation:** `../roadmaps/INDICATOR_V9_MA_GAP_ROADMAP.md`

---

## 📊 Performance Comparison

### **Win Rate Evolution**
```
v5: 36.4% WR (baseline - poor)
v6: 48.0% WR (+11.6% improvement)
v8: 57.1% WR (+9.1% improvement)
v9: TBD (expected >60% with momentum filter)
```

### **Key Learnings from Archived Versions**

**From v5 → v6:**
- Basic signal detection insufficient
- Need trend confirmation (HalfTrend)
- Color-based bar confirmation helps

**From v6 → v8:**
- 0.6% flip threshold prevents whipsaw
- Momentum confirmation critical (not just price cross)
- Anti-whipsaw logic needed (MA spacing)

**From v8 → v9:**
- **CRITICAL:** Don't short oversold (RSI 25-35)
- Short momentum at tops (Position ≥60%)
- ADX ≥23 filters weak chop (avoid ADX 20.7 disasters)
- RSI 50+ SHORTs = 68.2% WR (best performance!)
- Quality thresholds should differ by direction (LONG 90, SHORT 80)

**From v10 attempt:**
- Parameter insensitivity = no edge
- Simpler systems often better than complex
- Bug-free simple logic > buggy complex logic
- Don't add complexity without proven edge

---

## 🔍 Finding Archived Information

**By Feature:**
- Old indicators → Check indicator version numbers (v5, v6, v7, v8, v10)
- Deprecated systems → Look for "REMOVED" or "DEPRECATED" in filenames
- Historical baselines → Compare performance metrics in docs

**By Date:**
- Pre-Nov 12, 2024 → v5 era
- Nov 12-18, 2024 → v6 era
- Nov 18-26, 2024 → v8 era
- Nov 26, 2024+ → v9 era (current)

**By Performance:**
- Worst: v5 (36.4% WR)
- Moderate: v6 (48% WR)
- Good: v8 (57.1% WR)
- Best: v9 (expected 60%+ WR)

---

## ⚠️ Using Archived Documentation

**DO:**
- ✅ Reference for historical context
- ✅ Learn from past mistakes
- ✅ Compare old vs new performance
- ✅ Understand why changes were made

**DON'T:**
- ❌ Implement archived features in production
- ❌ Assume archived docs are accurate for current system
- ❌ Use archived parameters without validation
- ❌ Copy code from archived implementations

---

## 📝 Archiving Process

**When to Archive:**
1. Feature completely replaced by newer system
2. Performance proven inferior to current
3. No longer maintained or updated
4. Kept only for historical reference

**How to Archive:**
1. Move file to `docs/archived/` directory
2. Add deprecation notice at top of file
3. Update references in active docs
4. Remove from main navigation
5. Keep in git history (never delete)

**Deprecation Notice Template:**
```markdown
# ⚠️ ARCHIVED - [Feature Name]

**Status:** DEPRECATED (Replaced by [Successor])
**Date Archived:** [Date]
**Reason:** [Why replaced/removed]

**For current system, see:** [Link to replacement docs]

---

[Original content below...]
```

---

## 🗑️ What Gets Deleted vs Archived

**Archived (Kept):**
- ✅ Historical indicator versions
- ✅ Deprecated but once-production features
- ✅ Performance baselines and comparisons
- ✅ Design decisions that didn't work out
- ✅ Learning experiences worth documenting

**Deleted (Removed):**
- ❌ Temporary test files
- ❌ Duplicate documentation
- ❌ Incomplete drafts never used
- ❌ Outdated troubleshooting for removed systems
- ❌ Dead code with no historical value

---

## 📚 Historical Analysis Resources

**Performance Tracking:**
- Database: Trade table with `indicatorVersion` field
- SQL queries: `WHERE indicatorVersion IN ('v5', 'v6', 'v8')`
- Analytics UI: Compare tab shows version performance
- CSV exports: Historical backtest results

**Commit History:**
```bash
# Find when feature was archived
git log --all --full-history -- "docs/archived/filename.md"

# See original implementation
git show [commit_hash]:path/to/original/file.ts

# Compare versions
git diff [old_commit] [new_commit] -- path/to/file
```

**Documentation Trail:**
- Copilot instructions: Common Pitfalls section documents issues
- Deployment docs: Show when features were removed
- Roadmaps: Track feature evolution over time

---

## 🎓 Learning from History

### **Successful Patterns**
- **Iterative improvement:** v5 → v6 → v8 → v9 steady progress
- **Data-driven decisions:** Each change validated with trade data
- **Simple foundations:** v9 builds on v8 core, removes complexity
- **Direction-specific logic:** LONG and SHORT need different filters

### **Failed Patterns**
- **Premature complexity:** v10 added complexity without proven edge
- **Parameter insensitivity:** Identical results across configs = no edge
- **Rigid filters:** RSI <35 for SHORT caught wrong phase (oversold)
- **Ignoring direction differences:** Same thresholds for LONG/SHORT suboptimal

### **Key Lessons**
1. **Simple > Complex:** v9 removed RSI filter, improved performance
2. **Test at scale:** 72 v10 configs with identical P&L = clear signal
3. **Direction matters:** LONGs and SHORTs need different strategies
4. **Data beats theory:** RSI 50+ SHORTs = 68.2% WR (counterintuitive)
5. **Fix bugs quickly:** v10 removed after 1 day when bug discovered

---

See `../README.md` for overall documentation structure.