Look-Ahead Bias

When testing how a strategy would have performed historically, it is easy to accidentally use data that was only published later. This makes results look better than they could have been in reality.

Common causes include using company financial results as if they were known at the end of the quarter, when they were actually published weeks later; using economic data that were later revised, rather than the figures first released; and selecting companies based on information known only today, such as current index membership.

For example, a backtest that buys shares at the end of March based on first-quarter earnings would be using information not available until those earnings were reported in April or May, overstating the strategy's performance.

Look-ahead bias can make worthless strategies appear profitable. Avoiding it requires carefully tracking when information became available, using point-in-time databases that record data as it was originally published, and building in realistic delays. It is closely related to survivorship bias, which also distorts backtests.