DRILL Training · February 2026 · ~4 min
The five most common backtesting traps
A backtest is your strategy's mock exam — and it's alarmingly easy to turn it into an open-book one. Five traps almost every beginner falls into:
The five traps
- Look-ahead bias: the signal uses data that wasn't knowable at the time (deciding intraday actions with the close). Invincible in backtest, broken live;
- Overfitting: parameters tuned until history looks perfect — you fitted the noise, not the pattern. If small parameter changes wreck the results, it's unusable;
- Survivorship bias: testing only on instruments that are still alive, skipping the delisted and the zeroed;
- Ignored costs: fees, spread and slippage can eat the entire edge of a high-frequency idea;
- Short samples: tested on one bull leg only — the first range or bear market exposes it.
A simple self-check
Split the data: tune on the first 70% (in-sample), then validate untouched on the last 30% (out-of-sample). A sharp drop out-of-sample is the signature of overfitting.
Tip: passing a backtest is only gate one. Walk it manually with bar replay, then run it on a demo — only after all three gates does it deserve real money.