Point-in-time data shows a data set as it stood on each past date, including only what was known then and with the values originally reported. It removes the effect of later revisions, which would otherwise leak the future into a backtest.
Why ordinary data is not enough
Most databases show the latest version of each figure. A company's profit for a past year might have been restated, or an analyst estimate overwritten. A backtest using the latest version sees numbers that did not exist on the trade date. Point-in-time data stores each value with the dates it was first known and each time it changed.
An illustrative example
| Date known | Reported profit for the year |
|---|---|
| 15 May 2022 | Rs 500 crore (first reported) |
| 10 Feb 2023 | Rs 460 crore (restated) |
A test for June 2022 should use Rs 500 crore, because that is what investors saw. The restated Rs 460 crore only becomes usable after February 2023. The figures are illustrative.
What needs to be point-in-time
- Financial statements, dated by the announcement date and not the period end. In India, quarterly results are due within 45 days of the quarter end (60 days for the final quarter).
- Analyst estimates, which are revised often.
- Index and universe membership, as of each date. See stock universe selection.
- Macro data, which is often revised after first release.
- Prices adjusted for splits and bonuses, applied correctly for the date in question.
Why it matters
Without it, look-ahead bias is almost guaranteed. Point-in-time data is more costly and harder to obtain than ordinary data, but it is the foundation of a trustworthy backtest. It also helps with survivorship bias, since it keeps the companies that were listed at each date.
Frequently asked questions
What is point-in-time data?
It is data stored as it was known on each date, with the originally reported values and the dates of any later revisions, so a backtest sees only what an investor could have seen.
Why does point-in-time data matter for backtesting?
Because using revised or later-known figures leaks future information into the test and makes a strategy look better than it could have been in real time.
Do prices need to be point-in-time?
Raw prices generally do not change, but adjustments for splits, bonuses and other corporate actions must be applied correctly, and the universe of listed stocks must be correct for each date.
How do Indian reporting deadlines affect this?
Companies publish results some weeks after a period ends, within 45 days for quarters and 60 days for the final quarter, so figures should be dated by announcement and not by period end.