Price Series Discretization: Time Bars, Noise, and Event-Based Sampling
Summary
The article examines how time-based bars represent market prices and argues that fixed-interval sampling can distort a process whose changes are driven by trading and order-book activity rather than time itself. It contrasts the convenience, storage efficiency, visual scaling, and instrument comparison offered by time bars with the information they may discard. A Nyquist-style analogy illustrates how undersampling can obscure a signal, while the discussion of discrete prices, trades, and orders explains why market data do not map neatly to ordinary continuous-time signals.
As an alternative, the article proposes recording price changes after fixed price movements, creating a price-based rather than time-based scale. It argues this can reduce distortions associated with time sampling and make chart scaling more precise. The reasoning is conceptual and illustrated rather than supported by a systematic trading test. The author acknowledges that price formation has several possible drivers and that the underlying economic incentives cannot be fully measured; sampling choices should therefore be considered when interpreting patterns in any discretized series.
Key ideas
- Fixed-time bars are convenient and compact, but they can omit information about activity between sampling points.
- The article treats prices and traded quantities as discrete and emphasizes that elapsed time is not itself the cause of price changes.
- Undersampling is presented as a source of distortion that can make a structured signal appear noisy or random.
- Price-based sampling records observations after specified price moves instead of at regular time intervals.
- Patterns found in discretized data should be interpreted in light of the sampling method and its assumptions.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.