Skip to content

Trading System Strategy

Last Updated: 2025-11-18 Consolidates: 8 trading strategy and planning documents Status: Active - 2 live strategies (Micro-Cap, Phase 3D), 2 planned (Options, Crypto)

Executive Summary

Comprehensive strategy for building multi-agent AI trading system with automated strategy development, backtesting, paper testing, and live deployment.

Current State: 2 strategies live ($99K paper + $2.7K real capital), multi-agent architecture operational

Strategic Goal: 4-6 uncorrelated strategies generating 15%+ annual returns with <15% max drawdown

Strategic Pillars

1. Multi-Strategy Architecture

Active Strategies (2/4 deployed):

  1. Micro-Cap Momentum ✅ LIVE (Paper)
  2. Capital: $99K Alpaca paper account
  3. Status: 0/3 positions (waiting for signals)
  4. Strategy: Low-float micro-caps with unusual volume + momentum
  5. Entry: Price >$1, volume >1M shares/day, float <50M
  6. Exit: Trailing stop (ATR-based, 15-24%)
  7. Risk: Max $1K per position, max 20% allocated

  8. Performance Target: 25%+ annual, Sharpe >2.0, Win Rate >35%

  9. Phase 3D Financial Services ✅ LIVE (Real)

  10. Capital: $2.7K Schwab live account
  11. Status: 20 positions (dividend strategy)
  12. Strategy: Undervalued financial services stocks with dividends
  13. Entry: P/E <12, Dividend yield >3%, strong fundamentals
  14. Exit: Target 15% gain or fundamental deterioration
  15. Risk: Max 5% per position, diversified across 20 stocks
  16. Performance Target: 12%+ annual (capital gains + dividends)

Planned Strategies (2/4 in development):

  1. Options Theta Decay ⏳ Week 3 (Not Started)
  2. Capital: TBD ($10K allocation)
  3. Strategy: Sell cash-secured puts on quality stocks
  4. Target: 1-2% monthly income from option premium
  5. Risk: Max 2% loss per trade, only sell puts on stocks willing to own

  6. Timeline: Q1 2026 paper test, Q2 2026 live

  7. Crypto Trend-Following ⏳ Week 4 (Not Started)

  8. Capital: TBD ($5K allocation)
  9. Strategy: Follow major crypto trends (BTC, ETH)
  10. Entry: Moving average crossovers, volume confirmation
  11. Exit: Trailing stops, no overnight holds
  12. Risk: Max 3% per trade, max 10% allocated
  13. Timeline: Q1 2026 paper test, Q2 2026 live

Strategy Correlation: - Micro-cap vs Phase 3D: Low correlation (growth vs value, small vs mid-cap) - Options vs Equity: Low correlation (theta decay vs price appreciation) - Crypto vs Traditional: Low correlation (24/7 vs market hours, volatility) - Target: Portfolio correlation <0.3

2. Automated Strategy Development

Workflow: Idea → Backtest → Paper Test → Go/No-Go → Live Deployment

Phase 1: Backtesting - Tool: tools/strategy_tester.py - Data: Historical price/volume data - Metrics: Sharpe ratio, win rate, max drawdown, Calmar ratio - Pass Criteria: Sharpe >2.0, Win Rate >35%, Max DD <20%

Phase 2: Paper Testing (24-48 hour minimum) - Tool: Paper trader LaunchAgent (micro-cap example) - Execution: Real-time simulation with Alpaca paper account - Validation: Metrics match backtest (±10%), no critical bugs - Monitoring: /tmp/micro_cap_paper_trader.log

Phase 3: Go/No-Go Decision - Review: Backtest + paper test results - User Approval: Required for live deployment - Capital Allocation: Start small (10-20% of target allocation) - Monitoring: Daily P&L review, weekly strategy review

Phase 4: Live Deployment - Gradual scale-up (weeks 1-4) - Continuous monitoring and adjustment - Kill switch criteria (max daily loss: 2%, max weekly loss: 5%)

Tools: - tools/strategy_tester.py - Backtest engine - tools/trading_analytics.py - Performance analysis - trading_agents/ - Multi-agent trading system - .claude/skills/strategy-backtest/ - Comprehensive backtesting skill - .claude/skills/paper-test-strategy/ - End-to-end paper testing skill

3. Multi-Agent Trading Architecture

Agent Roles:

  1. Scanner Agent
  2. Scans markets for opportunities matching strategy criteria
  3. Filters candidates by liquidity, volatility, fundamentals

  4. Outputs: Watchlist of potential trades

  5. Analyst Agent

  6. Analyzes each candidate (technicals, fundamentals, sentiment)
  7. Scores each opportunity (0-100)

  8. Outputs: Ranked list with entry/exit levels

  9. Risk Manager Agent

  10. Validates position sizing, portfolio allocation, stop losses
  11. Checks correlation, exposure limits, margin requirements

  12. Outputs: Approved/rejected trades with risk parameters

  13. Execution Agent

  14. Submits orders to broker APIs (Alpaca, Schwab)
  15. Monitors fills, handles partial fills, implements stops

  16. Outputs: Executed trades with confirmation

  17. Monitor Agent

  18. Tracks active positions, P&L, portfolio metrics
  19. Triggers exits (stops, targets, time-based)
  20. Outputs: Alerts and trade exit signals

Communication: Agents coordinate via Redis (shared state) + direct API calls

Autonomy Levels: - Paper trading: Fully autonomous - Live trading <$1K position: Autonomous with alerts - Live trading >$1K position: Requires approval

4. Risk Management

Position-Level Risk: - Max position size: $1K (micro-cap), $2K (Phase 3D), $500 (crypto) - Stop losses: Required on ALL positions (ATR-based dynamic stops) - Position correlation: Max 0.5 between any two positions - Max concentration: 20% in any single strategy

Portfolio-Level Risk: - Daily loss limit: 2% of total capital - Weekly loss limit: 5% of total capital - Max drawdown: 15% (kill switch activated) - Margin usage: 0% (cash-only trading)

Capital Allocation: - Alpaca Paper: $99K (testing ground, no real money) - Alpaca Live: $971 (small positions, proven strategies only) - Schwab Live: $2,712 (primary account, conservative) - Total Real Capital: $3,683

Risk Controls: - Pre-trade risk checks (position sizing, stops, correlation) - Intra-day monitoring (P&L tracking, stop enforcement) - Post-trade analysis (lessons learned, strategy refinement)

Implementation Roadmap

Phase 1: Foundation (Complete ✅)

  • ✅ Micro-cap momentum deployed (paper)
  • ✅ Phase 3D financial services deployed (live)
  • ✅ Multi-agent architecture operational
  • ✅ Risk management framework implemented
  • ✅ Backtesting and analytics tools created

Phase 2: Strategy Expansion (Q1 2026)

  • [ ] Options theta decay strategy (backtest → paper → live)
  • [ ] Crypto trend-following strategy (backtest → paper → live)
  • [ ] Genetic optimizer for parameter tuning
  • [ ] Advanced backtesting with slippage/commissions

Phase 3: Optimization (Q2 2026)

  • [ ] Portfolio optimization (Kelly criterion, mean-variance)
  • [ ] Machine learning signal enhancement
  • [ ] Sentiment analysis integration
  • [ ] Alternative data sources (satellite, web scraping)

Phase 4: Scale (Q3 2026)

  • [ ] 6+ uncorrelated strategies operational
  • [ ] $50K+ capital deployed
  • [ ] Automated strategy discovery (AI-generated strategies)
  • [ ] Institutional-grade reporting and compliance

Success Metrics

Performance Targets

  • Annual Return: 15%+ (net of fees/commissions)
  • Sharpe Ratio: >2.0 (risk-adjusted returns)
  • Max Drawdown: <15% (peak-to-trough)
  • Win Rate: >40% (winning trades / total trades)
  • Calmar Ratio: >1.5 (return / max drawdown)

Strategy-Specific Targets

  • Micro-cap: 25%+ annual, Sharpe >2.0, Win Rate >35%
  • Phase 3D: 12%+ annual (8% capital gains + 4% dividends)
  • Options: 12-24% annual (1-2% monthly theta decay)
  • Crypto: 20%+ annual (higher volatility, higher returns)

Operational Metrics

  • Uptime: >99% (trading system availability)
  • Latency: <100ms (signal → order submission)
  • Accuracy: >99.5% (correct order execution)
  • Compliance: 100% (all trades within risk limits)

Architecture Decisions

Paper Test Before Live

Decision: Mandatory 24-48 hour paper test for all new strategies Rationale: Catches bugs, validates backtest assumptions, builds confidence Implementation: Alpaca paper account, real-time simulation, monitored logs Cost: 1-2 days delay, but prevents costly mistakes

Multi-Agent vs Monolithic

Decision: Multiple specialized agents vs single trading bot Rationale: Separation of concerns, easier debugging, more flexible Implementation: 5 agents (Scanner, Analyst, Risk Manager, Execution, Monitor) Trade-off: More complexity, but more robust and maintainable

Stop Loss Strategy: ATR-Based Dynamic

Decision: ATR (Average True Range) stops vs fixed percentage Rationale: Adapts to volatility, tighter stops in low-vol, wider in high-vol Implementation: 2x ATR for micro-cap (15-24%), 1.5x ATR for blue-chip (8-12%) Benefit: Prevents premature stops in volatile stocks, tighter risk in stable stocks

Correlation Limits

Decision: Max 0.3 portfolio correlation, max 0.5 position correlation Rationale: True diversification requires low correlation Implementation: Daily correlation matrix calculation, reject new trades if exceeds limit Benefit: Reduces portfolio drawdown, smoves returns curve

Consolidated From (archived in docs/archive/strategy-consolidation-2025-11-18/): - AGENTIC_TRADING_ROADMAP_2025-10-30.md (25.9 KB) - MULTI_STRATEGY_OPTIMIZATION_2025-10-31.md (17.1 KB) - genetic-strategy-optimizer.md (18.1 KB) - multi-strategy-channel-architecture.md (13.1 KB) - MICRO_CAP_DEPLOYMENT_READY_2025-10-31.md (13.2 KB) - PHASE2_PRACTICAL_GUIDE.md (6.1 KB) - + 2 more trading documents

Current References: - trading_agents/ - Multi-agent trading system - unified_api/ - Trading API abstraction (Alpaca, Schwab) - tools/strategy_tester.py - Backtest engine - tools/trading_analytics.py - Performance analysis - tools/portfolio_fetcher.py - Multi-account portfolio view - .claude/skills/strategy-backtest/ - Backtesting skill - .claude/skills/paper-test-strategy/ - Paper testing skill

Related Strategies: - MEMORY_KNOWLEDGE_STRATEGY.md - Trading history and lessons storage - ENTERPRISE_PLATFORM_STRATEGY.md - Overall IT Raven platform integration

Strategy Owner: Bert Frichot Review Cycle: Weekly (during active trading), Monthly (long-term strategy) Next Review: November 25, 2025 (weekly), December 18, 2025 (monthly)