Replies to post #10 on Advanced Automatic Investor

[DonCarlson]

Hi Conrad,

Yes, I had made the post about "Tweaking" on the wrong board (there's getting to be so many of these boards).

I primarily view this site as a forum to influence AI improvements. So, on the issue of AI portfolios, here's a couple weaknesses I see:
1. Individual parameters can not be established for each stock within a single portfolio.
2. If you do put multiple stocks in a portfolio, the Optimum module does not provide a way to automatically find the optimum parameters for the portfolio as a whole.

I agree that when you group several stocks together under one AI portfolio, and if you were able to find the best parameters for that particular group, you would for all practical purposes, be creating an index. The advantage, that I see, is that you would then have the best overall parameters for that group of stocks. Most portfolios with multiple stocks in the same portfolio will not be stocks from the same sector, therefore there is no generic parameters where one size fits all.

We all seem to more-or-less accept Lichello parameters, as a baseline, but it would be interesting if AI were able to optimize on a large list of stocks, and see how close Lichello was.

On your question of optimization based on weighted value of risk. To me, the answer to that is classic asset allocation, with annual rebalancing. Each index or ETF (AIMed seperately) would represent a different sector and be rebalanced on an annual basis, i.e.

The Madonna Portfolio (by Bernstein) would be:
10% S&P 500
10% U.S. small stocks
10% REITs
10% international large-cap stocks
10% international small-cap stocks
10% emerging markets stocks
10% precious metals stocks
30% U.S. short-term bonds

[aptus]

Hello Conrad,

Welcome to the board. I'm looking forward to hearing your ideas.

"This index will behave quite different than any one of the stocks and has also lower volatility than any one of the stocks."

The index will never have a lower volatility than all of the stocks (i.e. some stocks will have a lower volatility than the index and some will have a higher volatility), assuming all stocks in the index don't have identical volatilities.

I agree with you that AIMing each stock separately is the way to go. I don't see any advantage to AIMing a basket of stocks in one AI portfolio. The only reason you would do this is to diversify, however you can diversify by AIMing multiple stocks each in its own portfolio.

You can even go one step further and share the cash reserve amongst all of these stocks (i.e. AI thinks each stock has its own cash reserve, but the underlying brokerage account lumps all of the cash into one big pot). The danger here is that you might not be able to act on BUY recommendations in one portfolio if another portfolio has used up the cash.

I also have high hopes for using periodic rebalancing with AI. In the future, AI will include tools that will make this much easier. My current favorite approach is to use MPT (although I'm not out of the testing phase with this yet), but I'm keeping a eye on what Jibes comes up with (AIM-Rebal).

I'd also like to hear what you and others think about optimizing.

Regards,
Mark.

http://www.automaticinvestor.com