Key Takeaways
- • Trend-following filter research continues across time-series modeling approaches, with Part 10 building on prior analyses (Parts 7 and 9) examining digital signal processing techniques for technical trading applications.
- • First-order time series filters (MA, LWMA, ES) estimate constant mean values by attenuating noise in locally constant processes, while second-order filters (DMA, DLWMA, DES, SOMR) capture both mean and linear trend components in more complex price movements.
- • Second-order process modeling assumes inputs contain locally constant mean, linear trend component, and normally distributed random noise, requiring more sophisticated filtering approaches than simple moving average techniques for trend capture.
- • Digital signal processing frequency domain analysis provides quantitative framework for evaluating filter performance, previously documented in companion research examining filter efficacy across different market regimes.
Summary
This installment in the trend-following filters research series, published May 29, 2026, represents the tenth iteration of technical analysis methodology grounded in digital signal processing (DSP) principles. The article builds on foundational work presented in Parts 7 and 9, which examined common filtering techniques used by systematic traders to isolate trend signals from market noise.
The research distinguishes between two fundamental modeling approaches for time series processes. First-order filters—including moving average (MA), linear weighted moving average (LWMA), and exponential smoothing (ES)—operate under the assumption that input data contains a locally constant mean value contaminated by normally distributed random noise. These low-pass filters are designed to estimate the underlying mean by attenuating noise components, providing the baseline for trend identification in stationary or near-stationary price regimes.
Second-order filters represent an evolution in complexity, addressing price series that exhibit both a locally constant mean and an embedded linear trend component. This category includes double moving average (DMA), double linear weighted moving average (DLWMA), double exponential smoothing (DES), and the second-order process mean regression (SOMR) filter. By modeling the dual components of level and trend simultaneously, these filters capture directional movement more effectively than first-order methods, accommodating markets in transition or strong directional phases.
The methodological foundation relies on frequency domain analysis to evaluate filter performance characteristics, connecting the practical application of these techniques to rigorous signal processing theory. This quantitative framework enables systematic evaluation of how different filter architectures respond across varying market conditions and noise profiles. The series positions these filtering techniques as essential tools for algorithmic and systematic trading strategies seeking to distinguish genuine market trends from random price fluctuations.
The research distinguishes between two fundamental modeling approaches for time series processes. First-order filters—including moving average (MA), linear weighted moving average (LWMA), and exponential smoothing (ES)—operate under the assumption that input data contains a locally constant mean value contaminated by normally distributed random noise. These low-pass filters are designed to estimate the underlying mean by attenuating noise components, providing the baseline for trend identification in stationary or near-stationary price regimes.
Second-order filters represent an evolution in complexity, addressing price series that exhibit both a locally constant mean and an embedded linear trend component. This category includes double moving average (DMA), double linear weighted moving average (DLWMA), double exponential smoothing (DES), and the second-order process mean regression (SOMR) filter. By modeling the dual components of level and trend simultaneously, these filters capture directional movement more effectively than first-order methods, accommodating markets in transition or strong directional phases.
The methodological foundation relies on frequency domain analysis to evaluate filter performance characteristics, connecting the practical application of these techniques to rigorous signal processing theory. This quantitative framework enables systematic evaluation of how different filter architectures respond across varying market conditions and noise profiles. The series positions these filtering techniques as essential tools for algorithmic and systematic trading strategies seeking to distinguish genuine market trends from random price fluctuations.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.