LOADING
Skip to content
QUALITY UPDATES EVERY FRIDAY◭QUALITY UPDATES EVERY FRIDAY◭QUALITY UPDATES EVERY FRIDAY◭QUALITY UPDATES EVERY FRIDAY◭QUALITY UPDATES EVERY FRIDAY◭QUALITY UPDATES EVERY FRIDAY◭QUALITY UPDATES EVERY FRIDAY◭QUALITY UPDATES EVERY FRIDAY◭QUALITY UPDATES EVERY FRIDAY◭QUALITY UPDATES EVERY FRIDAY◭QUALITY UPDATES EVERY FRIDAY◭QUALITY UPDATES EVERY FRIDAY◭QUALITY UPDATES EVERY FRIDAY◭QUALITY UPDATES EVERY FRIDAY◭QUALITY UPDATES EVERY FRIDAY◭QUALITY UPDATES EVERY FRIDAY◭QUALITY UPDATES EVERY FRIDAY◭QUALITY UPDATES EVERY FRIDAY◭
Drift Engine
Trading TerminalFeaturesResearchPricingFAQ
Sign inStart for free
Home/Research/Testing & risk

Testing & risk

Backtesting without fooling yourself: look-ahead, overfitting and costs

A backtest answers “how would this fixed rule have done?”. Three mistakes make that answer look much better than reality.

Drift Engine Research7 November 20262 min read

On this page
  1. 01Look-ahead bias
  2. 02Overfitting
  3. 03Costs and execution
  4. 04In Drift Engine

Look-ahead bias

Using information that was not available at the time — a closing price before the close, a pattern derived from the same years it is tested on, revised data instead of first releases. The fix is discipline: every decision may only use what was known at that moment. Walk-forward testing builds this in.

Overfitting

With enough parameters, any history can be fitted perfectly — and the result will fail on new data. Fewer parameters, out-of-sample periods and results that hold across years and markets are the defences.

  • Per-year results. A strategy that made all its money in one year is a story about that year.
  • Parameter stability. Small changes in settings should not flip the result.
  • Honest selection. If a parameter was chosen on the test period, say so.

Costs and execution

Spreads, commissions, swaps and slippage add up, especially for short-term strategies. A backtest without costs is an upper bound, not an estimate.

TakeawayA good backtest tells you how a rule behaved — including the years it failed. A perfect one usually tells you how the rule was fitted.

In Drift Engine

The Backtest Engine tests seasonal patterns walk-forward, lead-lag between markets, spread mean reversion and combinations, reports every metric next to buy-and-hold, evaluates results per year and says openly where a parameter was chosen on the tested period.

Educational content. Not investment advice and not a recommendation to buy or sell any financial instrument. Trading currencies involves a high risk of loss.

Drift Engine

See it live for every currency

Start your first trial for free — 3 days of the complete terminal, every view open. Cancel before at no cost.

Start for freeSee all features

Keep reading

Testing & riskFrom a written strategy to a TradingView indicatorMost traders can describe their setup in a few sentences. Turning those sentences into exact rules is where vague ideas become testable — or fall apart.→Testing & riskMonte Carlo for traders: the drawdown your strategy has to surviveYour trade history is one path out of many possible ones. Reshuffling it thousands of times shows how bad the realistic bad case is.→Testing & riskProp firm challenges: the math behind passingProfit target, daily loss, maximum loss, minimum days, consistency — each rule changes how likely a given strategy is to pass.→
Drift Engine

Fundamental terminal for the currency market. Not investment advice, no signals.

ProductFeaturesPricingResearchGuidesFAQ
AccountSign upSign inReport a problem
LegalLegal noticeCreator programPrivacyTermsWithdrawalRisk notice
Contactcontact@drift-engine.netCancel contracts here
© 2026 Drift Engine. All rights reserved.Total price. As a small business under § 19 of the German VAT Act (UStG), we do not charge VAT.