Backtesting
Common Backtesting Mistakes
Avoid these pitfalls that can make your backtest results misleading.
Backtesting mistakes can make a bad strategy look good. Knowing these pitfalls helps you produce reliable, trustworthy results.
Common Pitfalls
- Overfitting — Adding too many conditions to perfectly fit historical data. The strategy looks great in backtests but fails in live trading
- Look-ahead bias — Using future data that wouldn't have been available at the time of the trade. Tradient's engine prevents this, but be cautious with custom indicators
- Survivorship bias — Only testing on instruments that still exist today, ignoring those that delisted or went bankrupt
- Ignoring transaction costs — Not accounting for spreads, commissions, and slippage can make an unprofitable strategy appear profitable
- Too short test period — Testing on only a few months of data doesn't capture different market conditions
- Cherry-picking dates — Choosing a test period that favors your strategy instead of testing across full market cycles
How to Avoid Them
- 1Keep strategies simple — fewer conditions, fewer parameters
- 2Always include realistic commission and slippage settings
- 3Test across at least 2-3 years of data covering different market conditions
- 4Use out-of-sample testing — test on a period you didn't optimize for
- 5Be suspicious of strategies with unrealistically high returns or win rates
If a strategy shows returns above 200% per year with less than 10% drawdown, something is probably wrong. Very few real-world strategies achieve those numbers.