No, do not assume this. I'm not comfortable selecting stocks that have a favorable RR balance and offering it as guidance simply because I've not tested it. It *appears* to have an edge, but I do not have any statistics to guide. From a gut-feeling point of view, and with a bit of knowledge about supply/demand, stocks with little overhead resistance should have very little supply (e.g., those folks who have been holding at those higher levels at a loss will be willing to unload the moment they are at the water line and even. Note this has a theme very close to Pascal's AB concepts.), and with little supply available, we should see price appreciation.Originally Posted by buzzman
The problem here is that we have a correlation issue. Simply put, some stocks react far more in line with the broader markets than do others. This makes the decomposition more difficult, e.g., it is harder to discern movement due to pivot resistance due to broad market movement.
My internal work is focused on principle component analysis, or PCA. Simply put, given a series of prices from several indexes as well as the equity, we should be able to start by removing, or detrending, the price series in a method that removes the influence of the index. Hence, if we start with the highest beta component, which generally is the equity price series, we can run the series through a correlation calculation to determine which index is most highly coordinated with the equity. Once we know this, we can apply PCA transforms to remove the influence of the index, and what remains (in principle) is now the stock behavior alone, without influences of markets. Note that it's impossible to do this in reality, because if you change the starting date of the series, you get a completely different set of weights which may be close to the previous set (or not). This is the part I'm struggling with right now. The implications are that two data sets, 500 days long, offset by 1 day, and run through this PCA mumbo-jumbo magic, do not produce curves that virtually overlap each other. This is the specific problem that I'm trying to understand. The magic works on canned data that I created, so I know I'm close, but not close enough.
Provided that I can overcome this pain-in-the-neck technicality, the next step is to recalculate all of the pivots using a detrending PCA algorithm. Basically, think of the pivots table that Maxime has constructed, but now do so without the influence of the markets. Once I have this, I can then test the assertion that stocks with little overhead resistance have a greater chance to move higher.
Regards,
Paul
Reply With Quote