I like to separate price change "signals" into a "spectrum" of components, that when added back together give the original price change signal. Some call this filtering or decomposition. The premise is that understanding some parts of price action accurately is preferable to understanding overall price action less accurately. Some components may trend well, or maintain a stable range (oscillate) while others may be chaotic and unpredictable. Why waste time trying to understand price components that have weaker patterns? One example is filtering daily price changes into the close-to-open change and intra-day change. Then you analyze each series independently. My research shows that the intra-day component trends better on a week-to-week time scale, over several years and several thousand stocks. You can also use volume, T&S and Level II data for filtering (qualifying) each price movement. These are just simple, obvious examples, not too powerful on their own, but definitely useful for charting or as the foundation of a higher level quantitative system. There are also more sophisticated filtering/decomposition methods that I have considered and tested. Does anyone want to discuss some ideas along these lines? How would you go about filtering price changes at a per-tick or per-minute time scale? Which price changes convey more information? Not every price change is equally important, right? Jay