Basic Real-Time Filters for Erroneous Tick and Quote Data

Summary

The document asks for research on tick and quote filtering methods with few tunable parameters. Its answer proposes a short set of basic validity checks: reject quotes where the bid is at or above the ask, flag trades printed well outside the quoted spread, verify that event timestamps are in sequence, and detect implausible maximum volume values.

The response argues that older filters may not transfer cleanly to current markets. Exchange technology and fragmented trade reporting can change which records appear anomalous, so filters designed around past market conditions may incorrectly discard valid observations or misclassify modern data artifacts. The suggestions are practical screening rules rather than a complete algorithm: the answer does not specify how to choose tolerances, handle exceptions, or validate filters against ground truth. It is best read as a starting checklist for real-time market-data quality control.

Key ideas

• Reject quotes when the bid price is at or above the ask price.
• Flag trade prices that fall unusually far outside the quoted bid-ask range.
• Check event timestamps for sequencing errors and screen for extreme volume values.
• Simple rules have few parameters but still require choices such as the permitted price tolerance.
• Historical filters may produce false positives as exchange systems and reporting practices change.

Tags

# Real-time Tick Filtering


# Real-time Tick Filtering












Is anybody aware of any papers regarding tick/quote filtering algorithms. I'm aware of the Olsen stuff, but I'd prefer something with fewer free parameters.

## Answer by madilyn (score 3)

https://quant.stackexchange.com/a/20868

You could try just the basics:

- Inversion of bid ask spread (`if bid_px >= ask_px`)

- Unusual prints far out (`if bid_px - eps_ticks <= trade_px <= ask_px + eps_ticks`)

- Time sequencing (`if event_time[0] >= event_time[1]`, `0` for most recent)

- Max values (`if volume == 2^64-1`)

There's very few papers out there and I believe all of them are outdated anyway. Some things that used to be positives the past would never happen given the exchange-side technology today and old quote filters would probably give you false positives over artefacts of the modern markets (e.g. fragmented trade reporting).

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.