Using a Kalman Filter as a Regime Check for Forex Mean Reversion

Summary

The article explains the Kalman filter as a recursive estimator that alternates between predicting a time-series state and updating that estimate with new observations. It describes how process and measurement variance affect responsiveness and smoothness, and notes limits including parameter sensitivity, computational cost, and delayed reaction to abrupt changes.

For its forex example, the article applies Bollinger Bands to a mean-reversion strategy and compares a baseline with versions filtered by a long-period exponential moving average or a Kalman estimate. Trades are considered when price moves beyond a band, with exits near the middle band and a fixed percentage stop. The article reports that the Kalman and moving-average variants selected different trades, with 71 overlapping trades, and describes backtesting and market comparisons. The supplied excerpt does not include enough complete performance results to establish whether either filter improves returns; the single-position rule also makes trade counts differ between variants.

Key ideas

• The Kalman filter recursively updates a state estimate using predictions and new observations.
• Process and measurement variance govern how reactive or smooth the estimate becomes.
• The example combines Bollinger Band extremes with a Kalman estimate as a confirmation filter for forex mean reversion.
• The article compares Kalman and moving-average filters and reports 71 shared trades between them.
• The available evidence does not establish that the Kalman-filtered strategy is more profitable or robust.

Tags

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