Adaptive Band-Pass Filtering with Rolling Autocorrelation

Summary

This indicator uses rolling autocorrelation to estimate a dominant cycle and tune the center period of a band-pass filter. It first applies a high-pass filter to reduce slower components, then evaluates autocorrelation across lags in a rolling window. The lag with the lowest correlation is doubled to estimate the cycle length. To keep the filter from shifting abruptly, the estimate is limited to a change of two period units per bar before it tunes the band-pass filter.

The indicator can display the filtered input, minimum correlation, estimated dominant cycle, or tuned output. The accompanying explanation suggests using changes in the tuned filter to locate possible turning points: a zero rate of change with a positive or negative filter value may indicate a nearby local high or low. This is an indicator concept rather than a tested trading system; the document gives no performance evidence. The cycle estimate depends on the chosen window and filtered price behavior, and the proposed turning-point interpretation should be treated as a signal for analysis rather than proof of reversals.

Key ideas

• A high-pass filter reduces slower components before autocorrelation is measured.
• The lowest-correlation lag is doubled to estimate the dominant cycle length.
• The cycle estimate is limited to small stepwise changes before tuning the band-pass filter.
• The indicator exposes the filtered series, minimum correlation, cycle estimate, and tuned output.
• A zero change in the tuned output may help flag possible local extrema, but no performance test is provided.

Tags

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