Adaptive Dual Kalman Filters Using Fractional EMA Inputs

Summary

This indicator combines a double fractional-length exponential moving average measurement with two single-state Kalman filters, one baseline and one faster variant. Each filter updates its price estimate using a gain based on state uncertainty, process noise, and measurement noise. It estimates noise from the squared difference between the EMA input and prior filter state, normalized by ATR; a short lookback informs process noise while a longer one informs measurement noise. A floor on measurement noise limits how small that estimate can become.

The chart plots both smoothed estimates and colors the space between them according to which is higher, offering a visual comparison of their relative movement. The document argues that this design can suppress sideways fluctuations while responding to structural shifts, but supplies no backtest, quantified comparison, or evidence that the plotted relationship predicts returns. The inputs and design rationale are experimental, so traders would need to assess behavior across instruments and settings before using it as a signal.

Key ideas

• A double fractional EMA transforms price before it enters the Kalman filters.
• Residual variance normalized by ATR informs adaptive process and measurement noise estimates.
• A measurement-noise floor is intended to prevent excessive confidence during quiet conditions.
• The indicator plots baseline and faster filter estimates and colors the area between them.
• The document offers a design rationale but no quantified evidence of predictive performance.

Tags

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