Point-in-time data
How do you judge a week honestly, using only the information that existed by the end of it?
In plain words
Point-in-time data means looking at a period strictly from the information available at that moment, with no later revisions, hindsight, or added context. Any measure that requires a longer record simply cannot be computed yet, and is left empty rather than guessed.
Why it matters
Markets are commonly described using series and rankings built over years. Judging an early period by standards that did not yet exist overstates what a trader could have known, and makes risk decisions look more informed than they actually were.
In a real week: ETH, week of 6 August 2018
ETH opened the week of 6 August 2018 at 408.77 and closed at 317.89, a fall of 22.23%, with a range of 35.75%. Its price record was only 25 weeks long, so no return percentile versus history could be computed, and no funding rate was available for the week. Any conclusion about whether this week was unusually bad had to be drawn without a longer comparison.
A common mistake
Filling in a missing statistic from a later, longer data set. A number that was not calculable at the time was not known at the time, and using it changes the question being answered.