The term look-ahead bias may mean:
- multi-period sampling bias,
- the inclusion of data that was not available at the time when back-testing, or,
- look-head benchmark bias.
Look-ahead benchmark bias
Look-ahead benchmark bias typically occurs when the composition of a benchmark index is used in a model or calculation without correcting for changes in its composition over time. This usually means using its end of period composition for the whole period.
From another point of view, this is the arguable inaccuracy of indices as a result of changes in composition: most major indices have market cap criteria for inclusion which leads to bad performers being dropped and replaced with good performers. It can be surprisingly large. This is a selection bias, but is very similar to survivorship bias.