Intraday Volume Forecasting with Seasonal Curves and Point Processes
Summary
The document compares approaches to forecasting intraday trading volume, with particular interest in futures. It highlights recurring patterns linked to time of day, contract rolls, and economic announcements. One approach models trades or volume through a doubly stochastic binomial point process, which can describe changing arrival intensity and support forecast intervals. It also points to examining volume curves across markets and activity during opening auctions.
A second perspective favors a simpler baseline: estimate a static intraday volume curve from recent historical data, optionally smoothing it with an exponentially weighted moving average. The contributor reports that this performed adequately for practical execution and did not see a significant forecasting gain from a dynamic model. That assessment is experience-based rather than a quantified comparative result in the text. Exceptional macroeconomic sessions may need separate treatment, and the discussion’s cited static VWAP analysis concerns equities, so applying it to futures requires judgment.
Key ideas
- Intraday volume patterns vary with clock time, contract rolls, and scheduled economic events.
- A doubly stochastic point process can model volume through changing event intensity and provide forecast intervals.
- A static curve estimated from recent intraday history and smoothed with an exponentially weighted average is a practical alternative.
- The document reports no significant forecasting improvement from dynamic modeling, based on one contributor’s experience.
- Macroeconomic sessions may need separate handling, and evidence cited for static VWAP focuses on equities.
Tags
Full text
# What are common methods for modeling intraday trading volume? # What are common methods for modeling intraday trading volume? What are the most common ways to model intraday trading volume, particularly for futures contracts? There are obviously a number of seasonal-type factors, like roll, economic news releases, time of day, etc. Are there any standard modeling techniques to deal with this type of problem, or any papers/books that are applicable? ## Answer by lehalle (score 9, accepted) https://quant.stackexchange.com/a/9504 The best paper is probably Relative Volume as a Doubly Stochastic Binomial Point Process - James Mcculloch. In this paper the volume is modelled via a Point Process, and theoretical laws are derived (with confident intervals, etc). And we put elements about this in Market Microstructure in Practice, Chap 2.1. Volume curves are analyzed, not only during the intraday auctions (comparing the volume curves over the world in GMT), but the intensity of order sent during the prefixing auctions. ## Answer by aajajim (score 4) https://quant.stackexchange.com/a/9506 From my point of view, dynamic models like the one developped in Relative Volume as a Doubly Stochastic Binomial Point Process - James Mcculloch to provide a dynamic forecast of the volume does not improve significantly the forecasting comparing to a static volume curve forecast using historical data (last month intraday data, and an EWMA algorithm). I've done some analysis of this kind in the past, and i can tell that from a practical point of vue (algorithmic execution), static forecast provide good results, you will have to deal with macroeconomic days separately! You can find a paper made by PragmaSecurities, Static VWAP: A Comparative Analysis, that deal with this subject for equities (but i think you can extrapolate it futures) : http://www.pragmatrading.com/sites/default/files/pdf/static_vwap_a_comparative_analysis_-_2009.pdf
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.