Annualizing Hourly Strategy Returns and Volatility

Summary

The document raises a backtesting question: how to report annual performance for a strategy evaluated on hourly open, high, low, and close data. The author has code that replays different portions of the historical series and calculates total return and total standard deviation, and asks whether those results can be annualized reliably.

No answer or calculation method is included, so the document does not establish a particular annualization formula or assess the size of estimation error. It highlights that hourly frequency alone is not enough to judge annual performance: the sample length, return dependence, and consistency of the strategy’s measurement period matter. In particular, scaling volatility by the square root of periods assumes suitable return behavior, while compounding returns requires a clearly defined return series and time horizon. The post is therefore a useful framing of a performance-measurement problem, but offers no empirical evidence or completed analysis.

Key ideas

• The question concerns annualizing returns and standard deviation from hourly strategy backtests.
• A meaningful annual estimate depends on the sample duration and the return series definition.
• Compounded return and volatility require different annualization calculations.
• Serial dependence can make simple volatility scaling unreliable.
• The document poses the problem but does not supply a method or test results.

Tags

# How to annualise hourly returns?


# How to annualise hourly returns?












I have hourly open,high,low,close candles data for a particular asset. I wrote my own algo and some back testing code that replays the data from the past hourly candles to calculate the total return and total standard deviation. So let's say my set contains 30k hourly candles, I can run my code and replay only first 100, 1k etc.

I'm thinking, how do I annualise the returns of this kind of data? Is it feasible at all or the "error" would be too big? The more general question is how to measure performance of a hourly strategy on an annual basis.

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.