Point-in-Time Market Data: Why It Matters and Where to Get It

Point-in-Time Market Data: Why It Matters and Where to Get It

If your backtests use data that was not available at the time, your results are fiction. Here is how to fix that.

What Point-in-Time Means

Point-in-time data is data that reflects only what was known at each historical moment, with no retroactive corrections, additions, or removals. It sounds like a basic requirement, but in practice, most market data sources violate it in ways that are difficult to detect and devastating to backtest integrity.

The most common violations are survivorship bias (your universe only includes stocks that still exist today, not ones that were delisted), lookahead bias (using data that was revised or corrected after the date in question), and retroactive universe construction (assuming a stock had options or met a liquidity threshold at a time when it may not have).

Why It Matters for Systematic Trading

The impact of point-in-time violations on backtest results is not marginal — it is often the difference between a strategy that appears profitable and one that is merely an artifact of data contamination. A momentum strategy that excludes delisted stocks will overestimate returns because the worst performers (which often delist) are removed from the sample. A short strategy backtested on today's micro-cap universe may include names that were not actually micro-cap at the time, distorting position sizing and universe composition.

Where to Get Point-in-Time Data

Databento

Databento's historical data is inherently point-in-time at the raw market data level — tick-by-tick records of what actually traded, when, and at what price. For price data, this is as clean as it gets. The limitation is that Databento provides raw data, not derived datasets, so any point-in-time universe construction must be done by the researcher.

Norgate Data

Norgate Data explicitly markets survivorship-bias-free historical data, including delisted securities. Their data includes adjusted prices, index constituents, and fundamental data. Norgate is particularly popular in the backtesting community for its clean handling of corporate actions and delistings.

Alphanume

Alphanume provides point-in-time data as a core design principle across all of its datasets. The historical optionable tickers dataset records which equities had listed options on each date — not which ones have options today. The historical market cap dataset provides capitalization as it was known at each date, capturing changes from share issuances, buybacks, and price movements. The dilution events feed timestamps filings as they were filed, with lifecycle tracking that is never retroactively altered. Every dataset is stored historically and remains fixed once published.

This approach is specifically designed for researchers who need to trust that their backtests reflect tradeable reality. The data does not change after the fact, which means your backtest from six months ago will produce the same results if re-run today.

Comparison Table

Making Point-in-Time a Priority

The practical advice is straightforward: treat point-in-time integrity as a non-negotiable requirement for any backtest you intend to act on with real capital. If your data source cannot guarantee that historical records reflect only what was known at the time, your results are suspect until proven otherwise. The providers listed above each address this requirement at different layers of the data stack.