Skip to content
All library documents

Choosing Sampling and Aggregation Methods for Sparse Tick Data

Article Quant Q&A · Author: Bazman
Read the original

Summary

The discussion considers how to turn exchange-provided half-second Level 1 buckets into hourly or daily observations. It treats fixed-time snapshots as a valid sampling choice, while emphasizing that the right price measure depends on the intended analysis. Alternatives include volume-weighted average price, a simple average, the last observed price, or the bid-ask midpoint.

For instruments that may go hours without trading, the answer does not prescribe a universal missing-data benchmark or averaging window. Instead, it says those choices depend on the downstream use of the data. It also questions estimating a daily close by averaging when a closing price is already available. The exchange is brief and does not compare methods empirically or specify handling rules for stale quotes, market closures, or different data objectives, so it serves as a framing of design choices rather than a complete aggregation protocol.

Key ideas

  • Fixed-time snapshots are a valid way to sample market data, but they may not suit every analysis.
  • The appropriate price measure depends on how the aggregated data will be used.
  • When no trade occurs, possible proxies include the last price or the bid-ask midpoint.
  • There is no universal missing-data rule or averaging period offered in the discussion.
  • If an observed closing price is available, estimating it through averaging may be unnecessary.

Tags

Market microstructureStatisticsExecution
Full text
# Aggregating Tick Data


# Aggregating Tick Data












I have Level 1 data that has already been aggregated into 0.5s buckets by the exchange.

I'd like to further aggregate the data into hourly and daily buckets. I plan to do this by simply taking a snapshot on each time interval i.e. at market open then market open plus one hour etc.

Q1.) Is there anything wrong in principle with this simple sampling strategy?

As always the devil is in the detail with these things. Some of the less liquid contacts may not trade for up to a couple of hours at a time what should I so in this situation? There will be several other issues I'm sure.

Q2.) Is there a set of benchmarks/best practice that someone can point me to for bucketing data where some points may be missing.

Q3.) Does it ever make sense to average some of the values (say over the last hour to estimate the daily close value? Is so are there any rules of thumb I can use to select the appropriate averaging period?

Thanks

## Answer by chollida (score 1)

https://quant.stackexchange.com/a/22917

> Q1.) Is there anything wrong in principle with this simple sampling strategy?

I mean sampling is a valid strategy, it just may not be the best. WOuld a VWAP style price be better? Would just an average be better?

Typically when no trade has happened you can model the price as the last, average of the bid/ask spread, etc. The price you want depends on what you plan on using it for.

> Q2.) Is there a set of benchmarks/best practice that someone can point me to for bucketing data where some points may be missing.

Again this can't be answered until we know what you want to use the data for.

> Q3.) Does it ever make sense to average some of the values (say over the last hour to estimate the daily close value? Is so are there any rules of thumb I can use to select the appropriate averaging period?

I don't understand this question. You already have a closing price included with your data, why try to estimate it?

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.