What is look-ahead bias?
The use of information in a backtest that would not have been available at the simulated decision time. Common examples: restated financials, revised analyst estimates, index membership computed retroactively. Look-ahead bias inflates apparent performance and is the most common methodological flaw in published backtests.
A backtest is supposed to simulate decision-making in real time. Look-ahead bias breaks the simulation by feeding the model information that wasn't available at the decision date. Example: a backtest of a value strategy uses 2023 annual earnings to make a January 2023 decision; but the 2023 annual was reported in March 2024. Using it in a January 2023 decision is impossible in real time; the bias inflates the apparent alpha. The correction is point-in-time data: every input must be lagged by its actual reporting delay.
How invest-like avoids it
The 7-framework consensus screen uses TTM (trailing twelve months) fundamentals as of the actual filing date, not the period-end date. A March-2024 filing of December-2023 results is treated as available from March 2024 forward, not from December 2023. The working paper documents the lagging methodology explicitly.
The /track-record/ page locks the entry timestamp at the date the screen actually surfaced the name, not retroactively at the next quarterly refresh. This ensures the realised returns reflect what an investor following the screen in real time would have experienced.
Frequently asked questions
What is look-ahead bias?
Using information in a backtest that wasn't available at the simulated decision time. Inflates apparent performance.
How does invest-like avoid it?
By using TTM fundamentals as of the actual filing date (not the period-end date) and locking entry timestamps at the date the screen actually surfaced each name.
How can I check for look-ahead bias in any backtest?
Look for explicit documentation of the lagging methodology, point-in-time data sourcing, and the use of fundamentals-as-of-filing-date rather than fundamentals-as-of-period-end.
Correo semanal gratis
Las acciones más sólidas de la semana, cada lunes.
Cada lunes: las acciones en las que la mayoría de los siete inversores coinciden por primera vez. Un correo, te das de baja cuando quieras.
Investigación, no asesoramiento de inversión.
Educational only. invest-like is not a registered investment adviser; nothing on this page constitutes personalised investment advice.