Trading Without Backtesting Isn't Strategy — It's Gambling

"This strategy looks great." So you put real money on it — and watched the account melt. Sound familiar?

I recently backtested an RSI(7) momentum strategy on gold's daily chart over roughly three years. The concept was counterintuitive — buy when RSI crosses above 70, sell when it drops below 30. The exact opposite of what textbooks recommend. And yet: $7,100 net profit, 61% win rate, profit factor of 2.0.

But the real value wasn't the profit number. It was the weaknesses the backtest exposed — things I never would have discovered without losing real money first.

1. Cap Position Size at 2% of Capital

Out of 18 trades in the test, 12 exposed more than 2% of capital to risk. The analytics tool flagged this as "critically high."

The result: a 33% max drawdown. A $10,000 account dropping to $6,650. Even though the strategy was net profitable, very few traders can psychologically survive that kind of decline without abandoning the system.

Fixed-fraction position sizing — limiting risk to 1.5-2% of total capital per trade — reduces absolute return size but dramatically improves survival odds.

2. Stop-Losses Aren't Optional

The original strategy had no stops. Positions stayed open until the opposite RSI signal fired. This is what enabled the "monster wins" — but it's also what inflated the drawdown.

The backtest optimizer produced an interesting finding: adding appropriate stop-loss and take-profit levels would actually improve net returns. By preventing outsized losses, even if you clip some winners, the overall expected value increases.

For a volatile asset like gold, ATR-based stops work well. A 2x ATR stop, for example, won't get triggered by normal daily noise but will pull you out of genuine adverse moves.

3. Use Time-Based Exits to Reduce Opportunity Cost

The median trade duration was 30 days. But some positions lingered for months and ultimately closed at small losses or breakeven.

Adding a time-based exit rule — say, 45 or 60 days maximum holding period — prevents capital from being trapped in directionless positions. That freed-up capital can be deployed on the next opportunity.

4. Analyze Long vs. Short Performance Separately

One interesting finding: long positions consistently outperformed shorts throughout the backtest.

This partly reflects gold's generally bullish trajectory during 2022-2025. But recognizing this asymmetry enables adjustments. For example, sizing short positions at half the size of longs, or requiring additional confirmation filters for short entries.

5. The Real Value of Backtesting Is Doubt, Not Confidence

This is the final and most important point.

You don't backtest to prove a strategy works. You backtest to learn where it breaks.

Without backtesting this RSI strategy, the most likely outcome would have been abandoning it during the early losing streak. But having run three years of data in advance, I knew the initial drawdown was within the strategy's "normal operating range."

Backtesting isn't a tool for validating strategies. It's a tool for mapping weaknesses and designing risk management that can handle them. Entering live markets without this process isn't trading — it's gambling.

FAQ

Q: If a backtest shows good results, will live trading produce the same? A: No. A backtest simulates past data, not future performance. Slippage, spread variation, and liquidity gaps all degrade real-world results. Think of backtest results as a ceiling estimate rather than a guarantee.

Q: How much data is enough for a meaningful backtest? A: Minimum 2-3 years. Ideally 5+ years covering different market environments — bull markets, bear markets, and sideways chop. This test covered approximately three years including both trending and ranging periods, meeting the minimum threshold.