Skip to content

03_voting_system.md

Latest commit

 

History

History
175 lines (130 loc) · 6.58 KB

03_voting_system.md

File metadata and controls

175 lines (130 loc) · 6.58 KB
Raw

Multi-Indicator Voting System Architecture

Target: Transform single MACD system to 90% accuracy ensemble
Implementation: Issues #250, #277-289
Status: Simplified Design - Create directories as needed

System Folder Structure

src/
├── voting/                           # Core voting system (Issue #250)
│   └── __init__.py                   # Foundation for base_voting_strategy.py
│
├── agents/                           # Agent ensemble members
│   ├── sentiment_v0.py               # V0-V4 agents as ensemble voters
│   ├── sentiment_v1.py               # Enhanced for voting integration
│   ├── sentiment_v2.py               # Market fear → regime input
│   ├── sentiment_v3.py               # Heuristic ensemble member
│   ├── sentiment_v4.py               # LLM reasoning member
│   └── tech_agent.py                 # Multi-indicator support
│
├── tools/                            # Market data and tools
│   ├── data_sources/                 # Market data, news APIs
│   ├── cache/                        # Performance optimizations (90% improvement)
│   └── news_governor.py              # Smart API sampling
│
├── utils/                            # Utility functions
├── analysis/                         # Metrics system
├── validation/                       # V4 validation framework
└── deprecated/                       # V0-V4 system reference documentation

Progressive Directory Creation Strategy

Phase 1: Core Voting (Issue #250)

Only create when implementing:

src/voting/
├── base_voting_strategy.py          # Core voting interface
└── majority_voter.py                # Simple 3/5 majority voting

Phase 2: Add Indicators (Issues #277-279)

Create when adding RSI/Bollinger/Volume:

src/indicators/                       # Create when implementing #277
├── base_indicator.py                 # Common indicator interface
├── rsi_indicator.py                  # Issue #277
├── bollinger_indicator.py            # Issue #278  
└── volume_indicator.py               # Issue #279

Phase 3: Advanced Features (Issues #281-289)

Create directories as features are implemented:

  • src/regime/ - When implementing market regime detection (#284)
  • src/execution/ - When implementing order management (#287)
  • src/analytics/ - When enhancing metrics for ensemble (#280)

Phase 4: Production Integration (Issues #258-263)

Create when ready for live trading:

  • src/integration/ - Alpaca API, real-time pipeline
  • src/monitoring/ - Performance tracking, alerts
## Implementation Flow (Critical Path)

### Week 1: Foundation (Issues #250, #277-280)
  1. src/voting/base_voting_strategy.py # Core voting interface
  2. src/indicators/rsi_indicator.py # First additional indicator
  3. src/indicators/bollinger_indicator.py # Second additional indicator
  4. src/indicators/volume_indicator.py # Third additional indicator
  5. src/analytics/ensemble_metrics.py # Track performance
### Week 2: Intelligence (Issues #281-283)
  1. src/voting/weighted_voter.py # Confidence-based decisions
  2. src/utils/signal_strength.py # Granular scoring
  3. src/analytics/indicator_performance.py # Track individual accuracy
### Week 3: Adaptation (Issues #284-286)
  1. src/regime/regime_detector.py # Market regime detection
  2. src/regime/regime_adapter.py # Dynamic weight adjustment
  3. src/timeframe/timeframe_manager.py # Multi-timeframe confirmation
### Week 4: Production (Issues #287-289)
  1. src/execution/order_manager.py # GTC order system
  2. src/execution/risk_manager.py # Risk controls
  3. src/analytics/performance_monitor.py # Real-time monitoring
## Key Design Principles

### 1. Modular Architecture
- **Each indicator is independent** - can develop/test separately
- **Voting system is pluggable** - start simple, enhance incrementally
- **Regime detection is optional** - system works without it initially

### 2. Backward Compatibility
- **Keep V0-V4 agents** - they become voting ensemble members
- **Maintain cache system** - 90% performance improvement preserved
- **Existing tools unchanged** - data sources, news governor, etc.

### 3. Progressive Enhancement
- **Phase 1**: Majority voting - 3/5 indicators need to agree
- **Phase 2**: Confidence weighting - better indicators get more influence
- **Phase 3**: Regime adaptation - behavior changes with market conditions
- **Phase 4**: Production features - order management, risk controls

### 4. Testing Strategy
- **Individual indicator testing** - Each indicator tested standalone
- **Ensemble backtesting** - Compare ensemble vs individual performance
- **A/B testing** - Current system vs deprecated system comparison
- **Production validation** - Paper trading before live deployment

## Integration Points

### With Existing System
- **Cache Integration**: All indicators use existing cache framework
- **Data Pipeline**: Leverage existing market data and news tools
- **Metrics System**: Extend advanced metrics for ensemble tracking
- **V0-V4 Agents**: Become voting members in current architecture

### With Production Features
- **Alpaca Integration**: Order manager connects to Alpaca API (#258)
- **Real-time Pipeline**: Voting system receives live market data (#259)
- **Strategy Orchestration**: Multiple voting strategies run simultaneously (#260)

## Success Metrics

### Technical Milestones
- **Week 1**: Basic ensemble voting operational (>3 indicators)
- **Week 2**: Weighted confidence system improves Sharpe ratio
- **Week 3**: Market regime adaptation reduces drawdown
- **Week 4**: Production-ready with order management

### Performance Targets (Research-Based)
- **Ensemble Accuracy**: 70-90% (vs current ~60% single MACD)
- **Sharpe Ratio**: >1.0 (research achieved 1.43 with weighting)
- **Max Drawdown**: <20% (research showed 32% reduction)
- **Individual Indicator Win Rate**: Track each indicator's contribution

## Next Steps

1. **Create folder structure** - Set up additional directories as needed
2. **Start with Issue #250** - Implement base voting architecture
3. **Add RSI (#277)** - First additional indicator for ensemble
4. **Parallel development** - Bollinger Bands (#278) and Volume (#279)
5. **Iterative enhancement** - Add features incrementally, test continuously

---

*This structure supports systematic development from single MACD to 90% accuracy ensemble while preserving valuable existing components.*