Here is a theory. Assuming the price movement of an individual stock is generally affected by the larger market or industry movement. Therefore when applying technical analysis on individual stocks, it would make sense to filter out the movement from such market or industry leaving only the movement that is specific to that stock. For example, let GM(t) be the historical EOD prices for GM and SPY(t) be the historical EOD prices for SPY. Furthermore, lets denote GM'(t) be the percentage of daily price change for GM and SPY'(t) be the percentage of daily price change for SPY. Instead of applying technical analysis on GM'(t) one could use GM'(t) - SPY'(t) to filter out the market movement and get a better result. Now the question are: 1. Is the assumption reasonable? 2. Has anyone done anything like this before? and if yes what was the result?
You are right, I am treading this as a signal process problem. It won't be hard to apply various signal processing technique on the price, however I am having problem figuring out a way to prove or disapprove the hypothesis. In order to test the hypothesis, one would need a control group. In this case, it would be some kind of technical indicators that have some real edges