One of the most debated topics in systematic trading: how should you trail your stops? There's no universal answer, but there is data. I ran a comprehensive analysis comparing three popular trailing stop methods on my XAUUSD EMA crossover strategy, and the results were illuminating.
The Three Contestants
1. ATR Trailing Stop
The ATR (Average True Range) trailing stop adjusts dynamically based on market volatility. In high-volatility environments, the stop widens. In low-volatility environments, it tightens. This is the most adaptive of the three methods.
How it works: Once a trade is in profit, the stop is placed at
Price - (ATR × Multiplier) for long trades. The multiplier typically ranges from 1.5 to
3.0.
2. Percentage Trailing Stop
The simplest approach: trail the stop a fixed percentage behind the current price. If gold moves to $2,000 with a 2% trailing stop, the stop sits at $1,960. Simple, predictable, and easy to understand.
How it works: Stop = Price × (1 - Percentage) for long trades.
Tested at 1%, 2%, 3%, and 4% trailing distances.
3. Chandelier Exit
Named for its hanging nature (like a chandelier from the ceiling), this method trails the stop from the highest high (for longs) by a multiple of ATR. It combines the volatility-adaptive nature of ATR with the anchoring to recent price extremes.
How it works: Stop = Highest High (N periods) - (ATR × Multiplier).
Tested with lookback periods of 14, 22 combined with multipliers from 2.0 to 3.5.
Test Setup
The baseline strategy is a dual EMA crossover on XAUUSD with a fixed 4% take profit, which I've been running for several years. I tested each trailing stop method against this baseline across:
- Multiple EMA combinations (9/21, 12/26, 20/50, 50/200)
- Multiple timeframes (H1, H4, D1)
- 3 years of data (2023-2025) covering different market regimes
- Various parameter sets for each trailing stop method
Key Findings
Overall Winner: ATR Trailing Stop
The ATR trailing stop with a 2.5x multiplier produced the best risk-adjusted returns across most EMA combinations and timeframes. Key metrics:
- Best Sharpe Ratio improvement: +15% over the fixed TP baseline
- Max Drawdown reduction: -8% compared to baseline
- Win rate: Slightly lower (-3%), but average win size increased significantly (+22%)
The reason? ATR naturally adapts to gold's volatility regime. When gold is trending strongly, ATR expands and gives the trade room to breathe. When volatility compresses, the trailing stop tightens and locks in profits faster.
Percentage: Simple but Inflexible
The percentage trailing stop at 2-3% performed adequately but couldn't adapt to changing market conditions. It worked well during steady trends but got chopped up in volatile, ranging markets. The fixed nature is both its strength (simplicity) and weakness (rigidity).
Chandelier Exit: Best for Trend Following
Interestingly, the Chandelier Exit with a 22-period lookback and 3.0x ATR multiplier performed best on the D1 timeframe with wider EMA combinations (50/200). It captured the most profit per winning trade of any method — but also had the lowest win rate.
The takeaway: Chandelier is best suited for trend-following strategies on higher timeframes where you want to ride big moves all the way.
What I'm Using Now
Based on this analysis, I've updated my live EAs:
- H1/H4 strategies: ATR trailing stop with 2.5x multiplier
- D1 trend-following: Chandelier Exit (22, 3.0)
- Scalping EAs: Still using fixed TP — scalping exits need to be fast, not adaptive
The best trailing stop is the one that matches your strategy's personality. ATR works for adaptive strategies. Chandelier works for trend riding. Percentage works for simplicity. Test on YOUR data, on YOUR strategy.
Takeaways for Your Own Testing
- Always compare against a baseline. Without a baseline (like fixed TP), you can't know if the trailing stop is actually adding value.
- Test across multiple market regimes. A trailing stop that works in 2024 might fail in 2025. Use data that covers trending, ranging, and volatile periods.
- Risk-adjusted returns matter more than raw returns. A method that makes 10% more with 30% more drawdown isn't better. Look at Sharpe ratio, Sortino ratio, and max drawdown together.
- Don't over-optimize parameters. If your trailing stop only works with an ATR multiplier of exactly 2.37, it's curve-fitted. Look for wide parameter ranges that produce stable results.
I'll share more detailed data and possibly the backtesting scripts in a future post. Follow me on TradingView or Twitter/X for updates.