Meaning and examples
Look-ahead bias occurs when a historical test uses information before it was published or available. The reporting period date is not the same as the date when an investor could have learned its contents. A figure for the previous year, published later, cannot be treated without qualification as known at the beginning of the new year. A backtest remains a hypothetical study, not a record of executed trades.[27](#source-market-performance)
A hypothetical strategy that buys in January using annual earnings from a March report relies on knowledge that was unavailable at the time. Restated data and the future composition of an index can create a similar problem. Even a correct modern database may not retain the historical versions that are needed.
How to control it
In a working dataset, separate the data period, publication, receipt and use in the decision. Then specify realistic execution after the signal appears. If dates or versions cannot be reconstructed, disclose the limitation instead of inserting a convenient point in time. Removing look-ahead bias does not eliminate costs, survivorship bias or the risk of overfitting the strategy itself. Related reading: [chart literacy](/guides/chart-literacy/) and [portfolio risk](/guides/portfolio-risk/).