A.I.M. Users Bulletin Board

[Tim Reese]

Hi Rien; re AGD/RSD

I am afraid that the stocks that produce good returns might not have much in common, at least superficially. Yes, that's possible, although I think you have to measure something otherwise you're making decisions on intuition alone.

For example, say a certain stock produces 40% returns, then it might well be that a minute change in the chart can change this result by 10% or more. This was in fact why I want to be able to produce AGD without any random functions in it. I think we're coming at this from different directions. My first inclination would be to generate lots and lots of different AGD sets in groups with different AIM or variability settings (like a Monte Carlo simulation). This would give me a distribution of results for every setting that indicates the range of returns expected. You then look to see if there are any obvious trends. Another approach would be to try to characterize the variability of stocks that give good AIM results, and then do a Monte Carlo across the successful range, looking for trends. This is what I would call data exploration and is not guaranteed to produce useful results.

I think that you are more inclined to test the patterns that you believe that you see in the stock charts. That's fine by me, but I'm not convinced that the chart patterns are very useful (prove me wrong, please) or that you can reduce the visual perception in the charts to something that's a reliable predictor.

But I will be able to enhance my subjective understanding, which should allow me to make a better guess at which stocks to AIM and with which parameters. IMO you can't make an intelligent selection without knowing the odds. The problem with testing patterns: you're assuming that the patterns you pick are representative, and you're taking an unknown risk that they are not. Choosing inputs at random yields an unbiased and therefore more representative sample.

Of course it is also possible that the result will be "just throw darts". There are indicators that this outcome could be rather likely. Nah. You can always incrementally improve your odds of a favorable result. Rather, does the effort required to achieve that increment bring sufficient payoff?

best regards Tim