ETF/Index Strategy Plan - 2025-11-01¶

Date: November 1, 2025 Category: ETF/Index Trading (Separate from Individual Stocks) Status: Planning Phase

🎯 Objective¶

Create separate ETF/Index strategies distinct from individual stock strategies. ETFs have different characteristics: - Lower volatility than individual stocks - Broader diversification within each position - Different holding periods (often longer) - Lower correlation to individual stock moves

📊 Target ETFs/Indexes¶

Technology/Growth Indexes¶

QQQ - Invesco QQQ Trust (Nasdaq-100 Index)
Tracks 100 largest non-financial Nasdaq companies
Heavy tech/growth exposure
High liquidity
VGT - Vanguard Information Technology ETF
Focused on tech sector specifically
Lower expense ratio than QQQ
More concentrated tech exposure
VTI - Vanguard Total Stock Market ETF (if VDI was typo)
Entire U.S. stock market
Maximum diversification
Lower volatility than QQQ/VGT

Innovation/Thematic¶

ARKK - ARK Innovation ETF
Active management (Cathie Wood)
Disruptive innovation focus
Higher volatility (more like individual stocks)
Note: Already included in "Volatile Stocks" strategy

🔬 Strategy Design Considerations¶

Why ETFs Need Different Strategies¶

Volatility Differences: - Individual stocks (TSLA, GME): Daily moves of 5-15% - ETFs (QQQ, VGT): Daily moves of 1-3% - ARKK exception: Daily moves of 3-8% (hybrid behavior)

Strategy Implications: 1. Mean Reversion Parameters: - Wider Bollinger Bands (2.5-3.0 std vs. 2.0-2.4 for stocks) - Lower RSI thresholds (25/75 vs. 27/63 for stocks) - Longer holding periods (signals less frequent)

Momentum Parameters:
Longer lookback periods (30-60 days vs. 10-20 for stocks)
Lower momentum thresholds (1-3% vs. 5-10% for stocks)
Position Sizing:
Larger positions OK (ETFs are already diversified)
$1,500-2,000 per position vs. $673 for individual stocks

📋 Proposed ETF Strategies¶

Strategy 1: ETF Mean Reversion (Conservative)¶

Symbols: QQQ, VGT, VTI Rationale: Low-volatility ETFs for mean reversion

Expected Characteristics: - Trade Frequency: 5-10 trades per year (vs. 18 for volatile stocks) - Win Rate: 65-70% (higher than individual stocks) - Sharpe: 0.8-1.2 (similar to volatile stocks) - Max Drawdown: 10-15% (lower than individual stocks)

Parameter Ranges (for optimization):

{
    "base_position_size": [1500, 2000],  # Larger positions
    "bb_std": [2.5, 3.5],                # Wider bands
    "max_positions": [2, 3],              # Fewer positions
    "rsi_overbought": [70, 80],          # More extreme
    "rsi_oversold": [20, 30],            # More extreme
    "stop_loss_pct": [0.08, 0.15]        # Tighter stops OK
}

Strategy 2: ETF Momentum (Trend Following)¶

Symbols: QQQ, VGT, ARKK Rationale: Ride tech sector trends

Expected Characteristics: - Trade Frequency: 8-15 trades per year - Win Rate: 50-55% (lower, but bigger wins) - Sharpe: 0.5-0.9 - Holding Period: 20-60 days (vs. 5-15 for stocks)

Parameter Ranges (for optimization):

{
    "lookback_period": [30, 60],         # Longer trends
    "momentum_threshold": [0.02, 0.05],  # 2-5% moves
    "position_size": [1500, 2000],
    "stop_loss_pct": [0.10, 0.18],
    "take_profit_pct": [0.15, 0.30]      # Ride trends longer
}

Strategy 3: ARKK Volatility Play (Hybrid)¶

Symbols: ARKK only Rationale: ARKK has individual-stock-like volatility

Expected Characteristics: - Trade Frequency: 12-20 trades per year - Win Rate: 55-60% - Sharpe: 0.6-1.0 - Behavior: Similar to volatile stocks (TSLA, GME)

Note: ARKK already performs well in "Volatile Stocks" strategy (502% return). This would be a dedicated ARKK-only strategy.

🔄 Optimization Plan¶

Phase 1: Data Collection (With Caching)¶

Before running experiments, implement data caching to avoid rate limits:

Pre-fetch all ETF data once:

# Create cache directory
mkdir -p data/cache/etf_bars/

# Fetch QQQ, VGT, VTI, ARKK data
# Save to data/cache/etf_bars/{symbol}_2018-2025.parquet

Modify optimization scripts to read from cache:

# In mean_reversion_optimized.py and micro_cap_momentum.py
# Add cache_dir parameter
# Check cache before calling Alpaca API

Time: 1 hour to implement + 30 minutes to fetch data

Phase 2: Run Optimizations (Sequential)¶

Launch experiments sequentially (not parallel) to avoid rate limits:

Experiment 1: QQQ/VGT/VTI Mean Reversion

uv run python3 tools/optimize_strategy.py \
  --strategy mean_reversion \
  --symbols QQQ,VGT,VTI \
  --train-start 2020-01-01 --train-end 2024-03-31 \
  --val-start 2024-04-01 --val-end 2025-10-30 \
  --population 20 --generations 20 \
  --param-ranges etf_mean_reversion \
  --output reports/etf_mean_reversion_conservative.json \
  --plot reports/pareto_etf_mean_rev.png

Wait 10 minutes, then:

Experiment 2: QQQ/VGT/ARKK Momentum

uv run python3 tools/optimize_strategy.py \
  --strategy etf_momentum \
  --symbols QQQ,VGT,ARKK \
  --train-start 2020-01-01 --train-end 2024-03-31 \
  --val-start 2024-04-01 --val-end 2025-10-30 \
  --population 20 --generations 20 \
  --output reports/etf_momentum.json \
  --plot reports/pareto_etf_momentum.png

Wait 10 minutes, then:

Experiment 3: ARKK Solo Volatility

uv run python3 tools/optimize_strategy.py \
  --strategy mean_reversion \
  --symbols ARKK \
  --train-start 2020-01-01 --train-end 2024-03-31 \
  --val-start 2024-04-01 --val-end 2025-10-30 \
  --population 20 --generations 20 \
  --output reports/arkk_volatility.json \
  --plot reports/pareto_arkk.png

Total Time: 3-4 hours (with wait times)

📊 Expected Results¶

Conservative Estimate (Based on ETF Characteristics)¶

Strategy	Symbols	Trades/Year	Sharpe	Return	Win Rate
ETF Mean Rev	QQQ, VGT, VTI	5-10	0.8-1.2	40-80%	65-70%
ETF Momentum	QQQ, VGT, ARKK	8-15	0.5-0.9	60-120%	50-55%
ARKK Solo	ARKK	12-20	0.6-1.0	80-150%	55-60%

Comparison to Individual Stock Strategies¶

Metric	Individual Stocks	ETFs (Expected)
Volatility	High (15-30% annual)	Low (10-20% annual)
Trade Frequency	15-36 trades/year	5-15 trades/year
Win Rate	53-61%	60-70% (more predictable)
Max Drawdown	20-30%	10-20% (lower risk)
Sharpe Ratio	0.19-1.08	0.5-1.2 (more consistent)

🎯 Capital Allocation Plan¶

Portfolio Structure (After All Optimizations Complete)¶

Total Capital: $99,000 (Alpaca Paper Account)

Individual Stock Strategies (50% - $49,500)¶

Volatile Stocks Mean Rev: $24,750 (25%)
Micro-Cap Momentum: $19,800 (20%)
Reserve: $4,950 (5%)

ETF/Index Strategies (40% - $39,600)¶

ETF Mean Reversion: $19,800 (20%) - QQQ, VGT, VTI
ETF Momentum: $14,850 (15%) - QQQ, VGT, ARKK
ARKK Solo (optional): $4,950 (5%)

Cash Reserve (10% - $9,900)¶

For rebalancing and opportunities

🔧 Implementation Requirements¶

1. Modify Optimization Scripts ✅ COMPLETED¶

ETF-specific parameter ranges have been implemented:

File: tools/mean_reversion_optimized.py (line 51)

Added asset_type parameter to constructor:

def __init__(self, symbols, initial_capital=10000.0, api_key=None, api_secret=None, asset_type='stock')

ETF parameter ranges in get_parameter_space() (when asset_type='etf'):

{
    "rsi_oversold": (20, 30),       # More extreme for low-volatility ETFs
    "rsi_overbought": (70, 80),     # More extreme for low-volatility ETFs
    "bb_std": (2.5, 3.5),           # Wider bands for lower volatility
    "stop_loss_pct": (0.08, 0.15),  # Slightly tighter stops
    "base_position_size": (1500, 2000),  # Larger positions (ETFs are diversified)
    "max_positions": (2, 3),        # Fewer positions (less frequent signals)
}

File: tools/optimize_strategy.py (line 76)

Added --asset-type CLI argument:

--asset-type {stock,etf}  # Affects parameter ranges (default: stock)

Verification Results:

Stock: bb_std (1.5, 3.0), position (300, 1000), max_pos (3, 8)
ETF:   bb_std (2.5, 3.5), position (1500, 2000), max_pos (2, 3)

2. Create Data Caching System ✅ COMPLETED¶

File: tools/data_cache.py (277 lines)

Features: - ✅ Fetch and cache stock/ETF bars from Alpaca API - ✅ Store in parquet format for efficient I/O - ✅ Rate limiting: 2-second delays between API calls - ✅ Cache management: list, clear, info commands - ✅ Lazy loading of API credentials from trading-command-center/.env

Usage:

# Fetch and cache ETF data (2018-2025)
python3 tools/data_cache.py --fetch QQQ VGT VTI ARKK

# List cached files
python3 tools/data_cache.py --list

# Show cache statistics
python3 tools/data_cache.py --info

# Clear cache for specific symbol
python3 tools/data_cache.py --clear QQQ

Cache Location: data/cache/etf_bars/

Cache Status:

Files: 4
Size: 0.43 MB
Symbols: ARKK, QQQ, VGT, VTI
Bars per symbol: 1968 (2018-01-01 to 2025-10-30)

3. Create ETF Channel Configs¶

Directory: trading_agents/config/channel_configs/etf/

Files to create: - etf_mean_reversion.yaml - etf_momentum.yaml - arkk_volatility.yaml

📅 Timeline¶

Week 1 (Nov 1-8)¶

[x] Document ETF strategy plan (this file)
[x] Implement data caching system (tools/data_cache.py)
[x] Pre-fetch QQQ, VGT, VTI, ARKK data (4 files, 0.43 MB, 1968 bars each)
[x] Modify optimization scripts for ETF parameters (--asset-type etf)
[ ] Run 3 ETF optimization experiments (sequential)

Week 2 (Nov 9-15)¶

[ ] Analyze ETF optimization results
[ ] Create ETF channel configs
[ ] Deploy best ETF strategy to paper trading
[ ] Monitor individual stock strategies (week 1 results)

Week 3 (Nov 16-22)¶

[ ] Deploy second ETF strategy if first validates
[ ] Continue monitoring all strategies
[ ] Begin options strategy design

🔍 Key Differences: ETF vs. Individual Stocks¶

Aspect	Individual Stocks	ETFs
Volatility	15-30% annual	10-20% annual
BB Std Dev	1.6-2.4	2.5-3.5
RSI Levels	27/63	20/75
Position Size	$673	$1,500-2,000
Max Positions	3-5	2-3
Trade Frequency	15-36/year	5-15/year
Holding Period	5-15 days	20-60 days
Win Rate	53-61%	60-70%
Risk Profile	Higher risk/reward	Lower risk/steady

✅ Success Criteria¶

For Each ETF Strategy¶

Minimum Thresholds: - ✅ Trades >= 5 per year (lower than stocks) - ✅ Win Rate >= 60% (higher than stocks) - ✅ Sharpe >= 0.5 - ✅ Max Drawdown <= 20%

Ideal Targets: - 🎯 Trades >= 10 per year - 🎯 Win Rate >= 65% - 🎯 Sharpe >= 0.8 - 🎯 Max Drawdown <= 15%

Current Strategies: - docs/status/OPTIMIZATION_ANALYSIS_2025-11-01.md (Individual stocks) - reports/mean_reversion_volatile_stocks.json (BEST individual stock strategy)

Architecture: - trading_agents/orchestrator/portfolio_manager.py - trading_agents/config/portfolio_config.yaml

Tools: - tools/genetic_optimizer.py - tools/optimize_strategy.py - tools/mean_reversion_optimized.py

To Be Created: - tools/data_cache.py (ETF data caching) - tools/etf_momentum.py (ETF momentum strategy) - trading_agents/config/channel_configs/etf/ (ETF configs)

Prepared by: Claude Code Date: 2025-11-01 Status: Planning Complete - Ready for Implementation Next Step: Implement data caching system before running experiments