What Is Point-in-time data?

Point-in-time data represents the information that was actually available at a particular historical decision time, rather than a later revised or completed version of the data. For each value used by a model or backtest, its source time, availability time, and trading decision time must be consistent: the value must be available no later than the decision it informs.

In quantitative research, point-in-time discipline applies to prices, index or trading-universe membership, fundamentals, macroeconomic releases, alternative data, and externally generated forecasts. It normally requires preserving historical versions, using an as-of join to select the latest value available at the decision time, and accounting for publication delays. Without it, a result may contain look-ahead bias even when its timestamps appear historical.

See also

• Decision-time availability

• Survivorship bias

• Look-ahead bias

Research and literature

• David H. Bailey, Jonathan M. Borwein, Marcos López de Prado, and Qiji Jim Zhu, Pseudo-Mathematics and Financial Charlatanism: The Effects of Backtest Overfitting on Out-of-Sample Performance, Notices of the American Mathematical Society 61(5), 2014.