A.I.M. Users Bulletin Board

[Tim Reese]

to Conrad et al. about testing

Conrad wrote

2) Although I think I know a thing or 2 about mathematics, maybe 3, the task completely eluded me and escaped between my fingers: I could neither formulate the input and the various parameters(Like Aptus noted), nor the output into any understandable or analysable structure which would lend itself to systematic analysis. I gave up for all I saw a chaos.

and

I do not think that such a formal in-depth attach will ever get off the ground, nor do I think that it will be worth the trouble.

I wouldn't mind being proved wrong on this.

I agree that the analysis has to promise some benefit to balance the effort. I think, with some good tools like Mathematica and a few little programs, one can generate a better picture of what's going on. I'm not necessarily proposing a closed-form, analytic solution - Monte Carlo immediately comes to mind as a likely technique. Certainly the problem needs to be posed in terms of summation rather than integration.

Also, the inputs are always going to be controversial. You could start off with backtesting stock data, then backtesting sectors, then maybe a means-reverting log-normal variant simulated signal. You don't need elegance - just a better intuitive understanding for what's going on and how the parts fit together.

A couple of things bug me, and set me off on this line of thought. First, http://www.jjjinvesting.com/chap12.htm the "Why Aim Works" chapter on Jeff's page, basically says to me not that the system works, but that it fails under some conditions. Here you have to put $35K at risk to rescue the stock... seems like a good chance to cut and run, or to reset the system as Tom suggests for deep divers.

Secondly, Myst's X_DEV http://stockwerld.com/xdev.htm claims to be an AIM derivative but doesn't give much insight on how it's different from AIM. Understandable, since Myst wants to sell the program, but intriguing nonetheless.

So, some kind of framework for testing and reduction of the problem could be useful, IMO.

Hope this isn't too much to digest in a single post...

best regards Tim