Hello Tim, Anybody,
I have copied the article you referred to but have not yet digested it fully.
In March this year I organised a test(This AIM Board) on the various AIM Structures that people had been using. With the Vortex AIM I can invoke several different modes and use different PC update algorithms. These(informal) test show a distinct sensitivity to optimisation(tweaking) of the parameters. We used an approximately random stock price wave pattern.
The results showed that there were significant differences between the models, indicating that for the given price structure some models were much better than others. All of them appeard to score higher than the basic Lichello AIM, inficating that various alternatives were indeed superior than the bare bones AIM By The Book. So I am somewhat puzzled that you say the testing you referred to gave disappinting results.
Look at messages # 1696, 1704 and 1754
If you are interested in receiving the Summary of these tests then give me you e-mail address and I will send it to you(if I can find it). Also, I have various Excel spread sheet of the Vortex AIM variants.
In regards to available stock data for testing AIMs I am sure there is. Various members have used such data but I do not have the sources. Someone will probably refer you to them.
The next stage of testing is what we are discussing now, but it should be done formally by experienced testers. In my opinion the objective is not so much testing the effectivities of different AIM Structures but the idea of automatic optimisation(various ideas can be incorporated into this goal. As I see it we should define exactly what it is we want to optimise and how we are going to do it, if we are going to do it. At the moment
I think we are still beating the bushes the get interesting people to sit up and notice and
participate. I hope you will stick around in this effort.
I am formulating my thoughts on this effort and in a few days I will make the ideas I have available in a structured format
Conrad
AI
