Trading Strategies

The system includes three built-in trading strategies, each designed for swing trading (holding positions for days to weeks, not intraday). All strategies comply with the PDT rule by using minimum holding periods of 2+ days.

The Three Built-In Strategies

1. Swing Momentum

Follows trends using moving average crossovers.

What it looks for:

• Fast moving average (10-day) crosses above slow moving average (50-day) = uptrend starting
• RSI confirmation: not overbought (RSI < 70)
• Rides the momentum until trend reverses

Buy signal:

• Fast MA crosses above slow MA
• RSI < 70 (not overbought)

Sell signal:

• Fast MA crosses below slow MA (trend reversal), OR
• RSI > 80 (severely overbought), OR
• 5% stop-loss triggered

Time horizon: 2-14 days (swing trades)

Best for: Trending markets, stocks with clear momentum

2. Mean Reversion

Buys oversold stocks, sells when they return to normal.

What it looks for:

• Price drops to lower Bollinger Band (2 standard deviations below 20-day average)
• RSI shows oversold conditions (RSI < 30)
• Bets on the stock bouncing back to its average price

Buy signal:

• Price touches or breaks below lower Bollinger Band
• RSI < 30 (oversold)

Sell signal:

• Price reaches middle Bollinger Band (mean), OR
• RSI > 60 (momentum returning to neutral), OR
• 5% stop-loss triggered

Time horizon: 2-10 days (short swing trades)

Best for: Range-bound markets, stocks with established support levels

3. Value Factor

Ranks stocks by fundamental value metrics.

What it looks for:

• Low P/E ratio (price relative to earnings)
• Low P/B ratio (price relative to book value)
• Low EV/EBITDA ratio (enterprise value relative to cash flow)
• Composite score ranks all stocks; buys top 20%

Buy signal:

• Stock ranks in top 20% by composite value score
• Rebalance occurs every 5+ days

Sell signal:

• Stock drops out of top 20%
• Rebalance period reached

Time horizon: 5-30 days (position trading)

Best for: Long-term value investing, fundamental-driven portfolios

How Strategies Are Ranked

The system uses walk-forward validation to test strategies on historical data, then ranks them by a composite score.

Walk-Forward Process

1. Train period: 24 months of historical data
2. Test period: 6 months of out-of-sample data
3. Step forward: Advance 3 months, retrain, test again
4. Repeat: Continue stepping through history

This simulates real-world conditions where the future is unknown.

Ranking Metrics

Each strategy is scored on five metrics:

Metric	Weight	What It Measures
Sharpe Ratio	30%	Risk-adjusted return (return per unit of volatility)
Sortino Ratio	20%	Downside risk-adjusted return (only penalizes downside volatility)
Max Drawdown (Inverse)	20%	Worst peak-to-trough loss (lower is better, inverted for scoring)
Profit Factor	15%	Gross profit ÷ gross loss
Consistency	15%	Percentage of positive months

Composite score formula:

score = (0.30 × Sharpe) + (0.20 × Sortino) + (0.20 × MaxDD_inv) +
        (0.15 × ProfitFactor) + (0.15 × Consistency)

Minimum Thresholds

To be eligible for live trading, a strategy must meet these thresholds in backtesting:

• Sharpe ratio: ≥ 1.0 (strong risk-adjusted returns)
• Profit factor: ≥ 1.5 (win $1.50 for every $1.00 lost)
• Max drawdown: ≤ 20% (largest loss no worse than 20%)
• Min trades: ≥ 100 trades (statistically significant sample)

Strategies that fail these thresholds are not deployed, even if their composite score is high.

Strategy Selection

The system displays ranked strategies on the Strategies page:

1. Top-ranked strategy: Highest composite score, meets all thresholds
2. Performance details: Sharpe, Sortino, drawdown, profit factor, win rate
3. Trade history: Recent signals and execution quality

You can:

• Enable/disable strategies
• View detailed backtest results
• Compare strategies side-by-side

ML Signal Strategy (Advanced)

In addition to the three rule-based strategies, the system includes an ML Signal Strategy that uses machine learning to predict stock movements. This strategy uses 50+ features and XGBoost to generate buy/sell signals.

See ML Signal Intelligence for details.

Configuration

Strategy thresholds and ranking weights are in config/settings.yaml:

strategy:
  min_sharpe_ratio: 1.0
  min_profit_factor: 1.5
  max_drawdown_pct: 20.0
  min_trades: 100

  walk_forward:
    train_months: 24
    test_months: 6
    step_months: 3

  ranking_weights:
    sharpe: 0.30
    sortino: 0.20
    max_drawdown_inverse: 0.20
    profit_factor: 0.15
    consistency: 0.15

Related Topics

• ML Signal Intelligence — Machine learning strategy
• PDT Rule — Why all strategies use min_hold_days ≥ 2
• Risk Management — Position sizing and stop-losses
• Glossary — Definitions of Sharpe ratio, Sortino ratio, etc.