Please advise how best to approach the following;
I have a trend following trading signal comprising a trailing simple moving average over n data points. The data is stationary. Currently i calculate a z-score based on the n data points, and judge the signal strength of the newest data point as the value of its z-score.
What alternatives might offer a better indicator of the strength of signal? It's worth noting that the data exhibits negative first order auto-correlation and (more difficult perhaps) the degree of negative auto-correlation is not constant - but let's say it varies in a predictable way.
Any pointers about what types of models i should perhaps be considering would be appreciated.