Handling Stale and Missing Data in Systematic Trading Systems
Summary
The document frames a trading algorithm as a system that combines prices, order book state, auxiliary data, information from related instruments, prior positions, and parameters. These inputs may arrive at different times, so a system needs to detect when observations are no longer current and decide how to act when values are missing.
It identifies possible responses including holding no position, forward-filling, extrapolating, temporarily reducing or increasing a data input’s weight, and inferring missing values. It also raises outliers and event-driven system design as related concerns. The section headings provide a useful map of data-quality decisions in live systems, but the supplied text contains no definitions, implementation details, examples, or evidence for choosing among the approaches. Any specific treatment therefore depends on context and is not established here.
Key ideas
- A trading algorithm may rely on inputs from multiple instruments and data sources with different update times.
- Stale observations and missing values require explicit handling decisions.
- Possible responses include forward-filling, extrapolation, changing an input’s weight, inference, or holding no position.
- Outliers and event-driven design are also identified as system concerns.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.