What is Point-in-time data?
Point-in-time data records each value as it was known on a given date, before any later revision. A GDP figure, an earnings number or a list of index members is stored as first published, with later changes kept as separate versions. Testing on point-in-time data stops a backtest from using numbers nobody had yet.
Why Point-in-time data matters
Many figures are revised after release, so testing on today's version quietly hands the past a look at the future.
Related terms
Lessons that use this term
- Look-ahead bias: when a backtest sees the future
- Survivorship bias: why testing only the winners lies