ATR Channel and Multi-Kernel Regression Trend-Following Strategy

Summary

This trend-following system combines an adaptive channel built from a simple moving average and ATR with a multi-kernel price estimate. The estimate blends Gaussian and Epanechnikov kernel regressions using configurable bandwidths and a weight. Crossovers between the regression line and the channel’s trend line generate signals, with RSI conditions filtering entries. Trade management uses a risk distance based on the channel and describes partial profit-taking and trailing exits.

The document gives parameter defaults and a published daily ETH/USDT backtest interval, but reports no performance metrics or evidence that the strategy is profitable. It warns that the result depends on ATR and kernel parameters, that smoothing can delay signals, and that ranging markets can produce false entries. It also notes the computational cost of kernel regression, especially for high-frequency use. Suggested extensions include adapting parameters to volatility and adding market regime filters; these are proposals rather than tested improvements.

Key ideas

• An ATR-width channel centered on a moving average updates its trend direction when price breaks through its boundaries.
• Gaussian and Epanechnikov kernel regressions are blended to estimate price movement.
• Crossovers between the kernel estimate and the channel line, filtered by RSI, define entries.
• Risk management uses channel-based risk distances, partial exits, and trailing stops.
• The published daily ETH/USDT test settings have no accompanying performance results, and the method may lag or fail in ranges.

Tags

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