Adaptive Cycle Filtering with Rolling Autocorrelation

Summary

The indicator estimates a market’s changing cycle length and uses that estimate to tune a band-pass filter. It first removes trend with a high-pass filter, then calculates rolling autocorrelations across a configurable range of lags. The lag with the lowest correlation is treated as half the dominant cycle, and the estimate is limited to moving by at most two bars per update before it sets the band-pass filter’s centre period.

The document describes four outputs and suggests using oscillator turns for timing, the cycle estimate to adapt other indicators, and the minimum correlation to judge whether cyclic structure is clear. It provides indicator settings and implementation details, but no performance backtest or evidence that these signals are profitable. The cycle estimate can be unreliable when the market has little cyclic structure, and its detected range depends on the chosen window.

Key ideas

• A high-pass filter removes trend before the indicator estimates cycle length.
• The most negative rolling autocorrelation lag is doubled to estimate the dominant cycle.
• The cycle estimate is smoothed by limiting its change to two bars per update.
• The resulting cycle estimate tunes the centre period of a band-pass oscillator.
• Cycle turns should be treated cautiously when the minimum correlation is near zero.

Tags

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