Kalman Filter Breakouts with Mean Absolute Error Bands

Summary

This strategy estimates a smoothed price and velocity with a two-state Kalman filter. It measures the recent mean absolute deviation of closing price from the estimate, then places adaptive bands around the filtered price. A close crossing above the upper band triggers a long entry; a cross below the lower band triggers a short entry. Opposite signals reverse the position, so the model can remain exposed as its directional signal changes.

The script exposes process and measurement noise, band lookback, and band width as parameters. The accompanying description says the values were tuned for ETH/USDT on a 15-minute chart and reports a backtest spanning more than 200,000 candles, but supplies no performance statistics or comparison. Its trend-following logic may behave differently across assets and timeframes, and the stated tuning is not evidence of out-of-sample robustness. The code declares an ATR trailing multiplier, but it does not use that input in its entries or exits; no explicit stop or profit target appears in the displayed strategy.

Key ideas

• A two-state Kalman filter estimates price and velocity from closing prices.
• Bands widen or narrow with the rolling mean absolute error around the filtered price.
• Crossings above or below the bands trigger long or short entries, with signals reversing exposure.
• The description reports ETH-focused parameter tuning and a large historical sample, without performance metrics.
• The displayed code has no explicit exit orders and leaves its ATR multiplier unused.

Tags

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