Look-ahead bias in EA testing: avoid using tomorrow’s information

A historical calculation can accidentally use information that was not available at the moment of a decision. The resulting chart may look convincing because the code is effectively seeing part of the future.

Attach a time to every feature

For each entry, identify which candle or event supplied the value and whether it had completed. A higher-timeframe candle may still be forming even though many lower-timeframe candles have already closed.

Inspect alignment boundaries

Check session changes, higher-timeframe boundaries and imported datasets. Avoid assigning a completed daily value to every minute earlier in that same day. Keep feature time and decision time in your diagnostic output.

Use a walk-through example

Pick one trade and reconstruct exactly what information the EA could access at that instant. Compare this with the values your analysis used after the period finished. This often reveals an alignment problem quickly.

Make the rule auditable

Readable source and explicit timestamps turn an impressive curve into an explainable process. Fixing a timing error may change the result substantially, but it gives you a much more useful foundation for future development.

Reference and further reading

Start with a free editable EA project · Explore MQLume source code

Back to top