7 Backtesting Mistakes That Destroy Trading Strategies

A strategy that returns 500% in backtesting can lose money live. The culprit? Backtesting mistakes that create false confidence. Here are the 7 deadliest errors and how to avoid them.

Mistake 1: Overfitting

Tweaking parameters until your backtest looks perfect creates a strategy optimized for past data that fails on new data.

Example:

Original strategy: Buy when RSI < 30
Backtest: 15% return
After tweaking: Buy when RSI < 28.3 AND price above 17-day MA
Backtest: 87% return (but fails live)

The second strategy is curve-fitted to historical noise.

Fix: Use walk-forward testing. Optimize on 70% of data, test on the remaining 30% (out-of-sample). If performance degrades >20%, you overfit.

Mistake 2: Look-Ahead Bias

Using information that wasn't available at the time of the trade.

Example:

if df['close'].shift(-1) > df['close']:
    buy()

Fix: Only use data available at the moment of decision. Use strict shift() rules and double-check your data alignment.

Mistake 3: Survivorship Bias

Backtesting only on coins that still exist excludes all the failed projects that would have wrecked your strategy.

Impact: Your "buy all altcoins" strategy looks great when you exclude the 80% that went to zero. Fix: Include delisted/dead coins in your backtest universe. Use historical exchange listings.

Mistake 4: Ignoring Fees and Slippage

Strategies with high trade frequency look profitable until you add 0.1% fees per trade and 0.2% slippage.

Backtest: 100 trades × 2% avg gain = 200% profit
Reality: 100 trades × (2% - 0.3% fees - 0.2% slip) = 150% profit
If average gain was 1%: 100% → 50% (half gone!)

Fix: Include realistic fees (0.05-0.1% maker, 0.1-0.2% taker) and slippage (0.1-0.5% depending on liquidity). See liquidity guide.

Mistake 5: Testing on Insufficient Data

Backtesting a strategy on 6 months of bull market data proves nothing about bear market performance.

Fix: Test across minimum one full cycle (2-3 years). Include:

• Bull market

• Bear market

• Sideways/choppy periods

• High volatility events (crashes)

Mistake 6: Unrealistic Position Sizing

Backtesting with "buy $10,000 of illiquid altcoin" when real market depth is $5,000 creates impossible results.

Fix: Check historical volume. Never backtest positions larger than 1% of daily volume. Include slippage that scales with position size.

Mistake 7: Not Accounting for Downtime

Backtests assume you're always watching the market. Real life has:

• Sleep (8 hours)

• Work (if not full-time trading)

• Sick days

• Exchange downtime

Fix: Build in realistic trading hours. If you can only trade 6pm-10pm, only test trades in that window.

Backtest Checklist

• [ ] Tested on out-of-sample data (walk-forward)

• [ ] Includes all delisted/failed coins (survivorship)

• [ ] No look-ahead bias in code

• [ ] Realistic fees (0.1-0.2% round-trip)

• [ ] Slippage modeled (0.2-0.5%)

• [ ] Minimum 2-year test period (full cycle)

• [ ] Position sizes match available liquidity

• [ ] Trading hours match reality

FAQ

How do I know if my backtest results are realistic?

If your strategy beats buy-and-hold by more than 2-3x, be skeptical. Most professional quant funds target 15-30% annual returns. If your backtest shows 200%+ per year, you likely have overfitting or data leakage. Check the strategy development guide.

Should I trust a strategy that works in backtest but fails in paper trading?

No. Paper trading is closer to reality than backtesting. If a strategy fails in paper trading, it will fail in live trading. Go back to your backtest and find the hidden bias (usually fees, slippage, or look-ahead).

---

Build reliable backtests with Trading Copilot's strategy workshop — built-in bias detection and walk-forward validation.