Using Singular Spectrum Analysis to Extract Price Trends

Summary

This document explains a price-trend indicator based on singular spectrum analysis, also called the Caterpillar method. It decomposes a recent price segment into additive components, then reconstructs a smoothed trend while filtering smaller-scale fluctuations. The author says the method does not require a stationary series, a predefined trend model, or known periodic components, and describes the indicator as avoiding the phase delay associated with conventional smoothing methods.

Users control the decomposition with parameters for history length, lag-window size, component count, and an optional noise threshold. The text explains how these settings affect smoothness and which fluctuations are retained. It advises using a relatively short history because computation grows costly, and gives practical parameter guidance and chart examples across multiple timeframes. These examples illustrate the indicator rather than establish trading performance. The article provides no systematic tests of predictive value, execution rules, or risk controls, so the extracted trend should not be treated as a validated trading signal.

Key ideas

• Singular spectrum analysis decomposes a recent price segment into components that can be recombined as a smoothed trend.
• The method is presented as usable without assuming stationarity or a known trend model.
• History length, lag-window size, and component settings control smoothness and noise filtering.
• Longer data segments raise computational costs, and the article recommends limiting the analyzed history.
• Chart examples illustrate the indicator but do not demonstrate profitability or predictive power.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.