A moving average of length n is just the sum of the previous prices multiplied by their weights. A normal moving average kernel gives all prices the same weight, an exponential ma kernel can be seen as a decay function , and a liner ma have a 1st order descending polynomial.
What I have been recently examining is using custom functions for a moving average kernel. Functions in the form of polynomials, windows, and piecewise smooth functions. What seems particularly interesting to me that I would like to ask others about is dynamic kernels. So essentially a moving average kernel that is allow to evolve as a response to price action. As of this point in time I found extremely curious crossover signals give by gaussian kernels that are modulated by the Kaufman's fractal efficacy ratio.
Has anyone researched custom moving average kernels or dynamic kernels to any extent? I would be very excited to read more on this but cannot find any relevant literature.
Kind regards
