Adjusting Continuous Futures Series at Intraday Frequency

Summary

The document raises practical questions about constructing continuous futures histories from minute-level data. It focuses on difference adjustment: how to choose the price difference at a roll, whether to use the final minute's close or a daily average, and whether the resulting offset should be applied across the entire preceding history. These choices affect the continuity and interpretation of intraday price series.

It also asks how to represent a roll spread gradually when a position is transferred over several days, and how roll handling should work in live data. The text offers no recommended convention, calculation steps, or empirical comparison. Its value is in identifying that daily roll procedures do not automatically specify intraday adjustments or live handling; the right approach depends on the intended series and roll process, which remain unspecified here.

Key ideas

• Continuous futures construction at minute frequency requires explicit roll conventions.
• Difference adjustment requires a defined price reference for measuring the contract spread.
• Historical offsets raise the question of which prior observations should be adjusted.
• A roll distributed across several days needs an intraday treatment consistent with its schedule.
• Live roll handling is raised as an operational question but not answered.

Tags

# futures roll - how is it applied to intraday data, e.g. minute level?


# futures roll - how is it applied to intraday data, e.g. minute level?












I have read many different approaches in rolling and adjusting futures contract to build a continuous time series. However, all of the examples I have seen are on daily data. While that is relatively straightforward to understand, how should I handle minute level data for example?

Take difference-adjustment for example - how is the difference determined? is it the close price of the final minute of the day the roll happens or average price of the day? then does this adjustment apply to all of the minutes in the entire historical data I have?

Further complication, if instead the roll happens over a number of days proportionally, rather than one single day - how do I handle that at minute level?

Thirdly, what happens when managing roll with live data?

Would be great if someone could talk me through the steps.

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.