I have been back testing some algorithms against a low volume highly volatile stock. I've found that during low volatile periods the technical indicators are following noise more than real trends. In order to filter out false signals I have been ignoring signals if volatility is below a fixed threshold.
How are flat low volume/volatility time periods usually handled?
EDIT: For a simple example: if the volume is low and there is not much price motion the EMA's will oscillate crossings creating false signals. In these cases small trade fluctuations ("noise") can create signals when there is no real trend present. While noise is obvious only in back testing, there seems to be a correlation between low-volatility/volume trading and a lack of a trend for stocks that are typically volatile when active.
As another idea I've been investigating rather than limiting signals by the volatility is to look at longer time averages for periods where trading is slow.
I'm interested to know if there is a more scientific approach to evaluating these type of situations or if there is a common methodology for optimization.