Backtesting

Evaluate rules on historical data with realistic assumptions.

Process

1. Define precise entry/exit/filters and management rules.
2. Use clean data; include fees/slippage; avoid peeking bias.
3. Segment by regimes; compute win rate, avg R, max drawdown, Sharpe.
4. Validate out‑of‑sample then forward‑test on paper before live.

Detailed Backtest Example: RSI Divergence Strategy

Strategy rules:

• Entry: Hidden bullish divergence on H1 when HTF (D1) is uptrend, RSI < 45, price touches 20‑EMA
• Stop: 25 pips below entry (or below prior swing low)
• Target: 2R (50 pips) for 50%, trail remainder with 20‑EMA
• Filter: Only trade London/New York sessions, avoid 30 min before major news

Backtest Results (EURUSD, Jan 2023–Dec 2023)

• Total trades: 87
• Wins: 48 (55.2%)
• Losses: 39 (44.8%)
• Avg win: +2.3R
• Avg loss: −1R
• Expectancy: (0.552 × 2.3R) − (0.448 × 1R) = 1.27R − 0.448R = +0.822R per trade
• Total return: +71.5R over 87 trades
• Max drawdown: −8.5R (consecutive 11 losses)
• Sharpe ratio: 1.42
• Largest win: +5.2R
• Largest loss: −1.2R (slippage)

Out‑of‑Sample Validation (Jan 2024–Jun 2024)

• Total trades: 42
• Expectancy: +0.76R per trade (slightly lower but acceptable)
• Win rate: 52.4% (similar to in‑sample)
• Conclusion: Strategy is robust; proceed to forward testing

Forward Test (Paper Trading, Jul 2024–Sep 2024)

• Total trades: 28
• Real expectancy: +0.68R (includes real slippage/spreads)
• Note: Slippage reduced expectancy by ~0.1R (expected)
• Decision: Go live with 0.5% risk per trade initially

Common Backtesting Mistakes

• Overfitting: Optimizing 10 parameters → 95% win rate (unrealistic)
• Solution: Limit to 2–3 parameters, test on multiple pairs/timeframes
• Ignoring fees: Backtest shows +100R, reality +85R after commissions
• Solution: Add 0.1% commission + 1 pip slippage per trade
• Survivorship bias: Testing only current liquid pairs
• Solution: Include pairs that were delisted/merged