Almost every misleading backtest comes down to letting hindsight, cherry-picking, or hidden costs creep in. Fix these nine and your test starts telling you the truth - which is the entire reason to run one.
1. Changing rules after seeing the outcome
The cardinal sin. You see price reverse, so you "would have" closed early or skipped the trade. In real time you had no such knowledge. Write your rules down before the test and follow them mechanically. This single discipline fixes more broken backtests than anything else.
2. Skipping losing trades
Quietly ignoring setups that "did not look clean" removes losers your rules would have taken, understating drawdown and inflating win rate. Take every trade that meets your criteria, ugly ones included.
3. Ignoring spread, commission, and slippage
Costs hit every real trade. A strategy with +0.05R of raw expectancy can go negative once a one-pip spread is subtracted per trade. Always test on results that include realistic costs.
4. Testing only one market condition
A strategy that shines in a trend can die in a range. If your sample is one calm quarter, you tested the market's mood, not the strategy. Span trending, ranging, and volatile periods - the heart of testing across market conditions.
5. Too small a sample
Twelve great trades prove nothing. A single outlier dominates small samples. Gather enough trades - usually 100 or more across conditions - before trusting any figure.
The same strategy looks five times better once two innocent shortcuts sneak in. Live trading then delivers the honest number.
6. Over-optimizing the parameters
Tuning every setting until the past looks perfect produces curve fitting - a strategy fitted to history that fails on new data. Prefer robust rules that work across a range of settings over a single magic combination.
7. Not writing the rules down
Vague rules invite hindsight because you can reinterpret them mid-test. Precise, written rules - exact entry trigger, stop, target, risk, session, and news filter - make the test repeatable and honest.
8. No forward test
Even a clean backtest can be lucky. Confirm the edge on data you have not seen through forward testing before risking real money.
9. Judging on profit instead of process
A profitable test built on broken rules teaches you nothing repeatable. Judge whether you followed the plan first, then whether the plan made money. Process is what transfers to live trading.
Important: the goal of a backtest is not a pretty number. It is an honest estimate of how a plan behaves. Every mistake above trades honesty for prettiness - and the market collects the difference later.
Backtest in a way that resists these mistakes
Bar-by-bar replay makes hindsight much harder because the next candle has not printed yet. When you use a simulator that hides the future, auto-sizes risk, applies costs, and logs every trade, most of these mistakes become difficult to commit by accident - which is exactly the point.
Backtesting mistakes FAQ
What is the biggest backtesting mistake?
Changing rules after seeing what price did next. Define entry, stop, target, risk, and filters in writing before the test and judge the data only after the sample is complete.
Why do my backtests look better than my live trading?
Usually ignored costs, cherry-picked trades, or curve fitting. Add realistic costs, take every valid signal, and forward test to close the gap.
Do I need to include spread and commission?
Yes. Costs are paid on every real trade and can flip a marginal strategy into a loser, so always include realistic spread and commission.