Choose Backtest Length by Statistical Evidence, Not a Fixed Time Span

Summary

The document addresses how far back a trading system should be tested. Its answer shifts the focus away from a standard number of months or years and toward whether the test produces enough observations to meet the statistical criteria chosen by the researcher. This frames backtest length as a question of evidential sufficiency rather than convention.

The response is brief and gives no specific criteria, sample-size method, or example for deciding when results are sufficient. It also does not discuss whether historical data cover different market regimes or how data quality and strategy horizon affect the choice. The main lesson is to define the evidence standard first and gather enough results to assess it; the note alone is not a complete validation procedure and does not prescribe a universal testing period.

Key ideas

• There is no single backtest duration that is appropriate for every trading system.
• The amount of data should be judged against the statistical criteria selected for evaluation.
• The response gives no specific sample-size rule or validation procedure.
• Market regimes, data quality, and strategy horizon are not addressed in the brief answer.

Tags

# How far back is normal to backtest an ATS ?


# How far back is normal to backtest an ATS ?












> Possible Duplicate: How much data is needed to validate a short-horizon trading strategy?

How far back do people usually backtest trading systems? months? years?

## Answer by Craig (score 0)

https://quant.stackexchange.com/a/3929

I think this question has been asked before, but it's not about time-frames per se, but about having enough results to satisfy whatever statistical criteria you have decided to adopt.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.