Replies to post #1159 on A.I.M. Users Bulletin Board (AIMUSERS)

[aptus]

Hello Conrad,

"Now, all other things being equal (which is never the case), if we start 2 AIMs and you make bigger buys that I do early in the game and I make bigger buys that you do later in the game then (eliminating other possibilities) there are two extremes that we consider..."

That assumes that the number of buys are equal -- which probably won't be the case. AIM might make 10 buys on the way down (before it runs out of cash) while Vortex only makes 5 (the first 2 Vortex buys might be smaller than AIM's first 2, but the next 3 might be significantly larger -- then Vortex runs out of cash while AIM goes on to buy 5 more times at a lower price).

Now, as you've mentioned, this all depends on how you set the buying behaviour in vortex. So you could set its behaviour such that it performs as in your last post. However I'd counter by saying that if I thought prices would continue to decline for a long while, I'd set my Buy Resistance (and/or minimum % per trade) differently to compensate.

What it comes down to is the system that wins will depend on how the declining price behaves and the respective tuning parameters.

Either system can be set to maximize returns for a particular price pattern by tuning the appropriate configuration parameter. Vortex would tune the "multiplier" parameter and AIM would tune the Buy Resistance. But since we don't know how the price will behave, neither of us can tune the relevant parameter exactly (if we could then we would know the price pattern and would have no need for AIM or Vortex).

So you're left with two systems that depend on a parameter (it also depends on the price behaviour, but you can't control that). My question then becomes, "if you have two systems that depend on a parameter, which one do you use?"

The obvious answer is that you go with the one with which you have the most confidence and/or familiarity. In order for me to go with some other algorithm I need to see it giving me a significant advantage over what I'm currently using. If all I've done is trade one parameter setting for another one, I don't see the benefit. Furthermore since I'm familiar with the AIM method and I've backtested it and know how it works (which I haven't done for the Vortex method), it makes sense for me to stick with that method.

If we backtrack for a moment and forget about tuning parameters, then Vortex will need to supply a "general case" multiplier. AIM would use whatever BR/SR/Min%perTrade settings were chosen when initially setting up the portfolio. In that case I believe that AIM would come out ahead most of the time. I suppose the only way to verify this is to backtest both algorithm using the testing methods I alluded to in a previous post.

To conclude, I think your idea can work, however I don't believe it has any advantage over AIM in the best case and might be at a disadvantage in the worst case.

Having said all of that, however, I think people should use whatever algorithm they think is best. After all, nobody else cares as much about your money as you do. If you think Vortex is the better algorithm then you should use it.

Whew! That was a much longer note than I had intended to write.

Regards,
Mark.

http://www.automaticinvestor.com

[labestul]

Hi Conrad,

I just visited your pages on the vortex method. I found them very interesting and I have the following comments:

(1) There are a number of "TYPOS" most of which are unimportant but some of which might be misleading to the casual reader. For example: In the section "The Multifunctional Case" just after you speak of the Lichello Flaw you give the example ...

y1=5000 y2=3000 pc1=5000
pc1-y2=5000-3000=2000

but then you say

Test1 --- Advice= Buy 3000 ...

Clearly this 3000 is incorrect and merely a typo because you subsequently correctly use 2000.

(2)Perhaps I am rehashing old ground here because I gather that this may have been discussed over at TMF. I plan to read all of the post over there but these days it's hard just keeping up over here. In essence what you term the Lichello Flaw is something that I first noticed with AIM and it bothered me terribly but mostly for esthetic reasons.

I think it is unfair to describe it as a flaw ... I think of it as a feature. Of course if one uses GTC orders then it can be considered a flaw but if one uses periodic updating (e.g. monthly) which is the context in which Lichello was operating, then it can be a blessing in a rapidly decreasing market.

(3) Your answer to the flaw is in effect to buy an amount greater than what standard AIM would suggest and that amount would be such that the revised portfolio control would equal the total value of the stock held after the buy. I played with this idea more than two years ago. I ran countless simulations and found that this revised method sometimes outperformed standard AIM and sometimes did not.

(4) Your idea of employing adjustments to portfolio control both with buys and sells (and not necessarily the same adjustments) together with the idea of possibly using negative values of "f" for example ... is a very interesting idea. I will perform some simulations in this area. One could of course claim that Lichello used two separate factors with AIM. One adds to the portfolio control 50% of the amount purchased and 0% of the amount sold.

(5) There is a significant error in at least one place. You use "S" to represent Lichello's SAFE-factor and you then state that Lichello's buy-advice is S*(pc1-y2). There are in fact two errors here. First Lichello's "buy advice" is actually (pc1-y2) regardless of the value of S which is 10%. The "buy order" is this "buy advice" modified by S. In other words your term "buy-advice" seems to correspond to Lichello's "buy order" rather than Lichello's "buy advice". OK ... OK this is not really an error but more a question of semantics. This might however confuse AIMers who were reading casually. Let's use your term of "buy-advice"

You state that Lichello's buy-advice would be

S*(pc1-y2)

but this is incorrect!

Lichello's buy-advice (i.e. buy order in Lichello terminology) is

pc1-y2*(1+S)

Your fourth summary point (i.e. By adjusting the f-actor downwards by a small amount this has an identical effect as using a SAFE-factor, as is done in the Lichello AIM Model. The use of the f-factor eliminates the need for a SAFE altogether as the investor can tune the investment aggression exactly to his needs. ) seems to be based on this incorrect equation. This is something however, that I will investigate further.

I hope the above comments are helpful. Thanks for provide a very interesting colour variation to AIM.

Barry