I tested all of the DOW stocks and most of the Naz 100. I also tested a wide variety of 2nd and 3rd tier Internet stocks (e.g. INKT, CMGI, ICGE) as well as some highly volatile issues (e.g. MSTR).
I usually used time periods starting from 1970 (sometimes I went back to the 1960s, but only in a few cases) and continuing to 2002 (or I used the first date that data was available for the stock if they weren't available in 1970).
In order to test, I did the following:
1. Start with the earliest available date and test for 5 year periods (e.g. for IBM, start with 1970 and test from 1970 to 75, then 1971 to 76, 72, to 77, etc.). I then repeated this process for 3 year and 10 year periods.
2. From 1990 to 2002 I performed similar tests, but this time using 1 year periods in addition to 3 year and 5 year periods.
3. I compared the results between the various Initial Cash settings as well as to buy and hold. For these tests I used the Standard lichello AIM settings (i.e. 50/50, 67/33, 80/20 with 5% minimum purchases -- I tried using 10% minimums, but that didn't work well so I stuck with the 5%).
For the most part, the lower the initial cash setting, the better the performance. One interesting note is that most DOW stocks had better Buy and Hold performance than the performance obtained by using AIM (with any initial cash setting). This leads me to believe that the majority of DOW stocks really aren't well suited for standard AIM.
When adding volume to the mix, however, AIM did much better with the DOW stocks (compared to standard AIM. I posted the results over at the SI board earlier this year), however Buy and Hold was still slightly ahead.
On the other hand, most of the Naz 100 stocks and Internet stocks significantly outperformed buy and hold with AIM using Volume. Some stocks (e.g. ICGE, MSTR) showed losses with Buy and Hold, but hundreds of percent gains with AIM + volume.
Others (e.g. YHOO, ORCL) showed gains with Buy and Hold but much larger gains with AIM + volume.
The lesson I took away was simply this, without proper testing you cannot be sure that AIM (standard or with Volume) will beat buy and hold for a particular stock (actually you can never be 100% sure since the past is no indication of the future, but at least you can have a better idea of what is probable assuming the stock's characteristics don't change).
To date there has not been an easy way to historically test stocks and funds. So most AIMers use an ad hoc, gut feeling strategy. This might work most of the time for the experienced people, but what's really needed is a step-by-step checklist on how to test stocks for AIM.
Testing involves much more than simply running a few stocks through a spreadsheet or an AIM software's historical analysis module and viewing the results. There should be a methodical method in place that allows you to properly test all stocks in the same way (i.e. over a wide variety of time periods and market conditions).
There are many methods that can be employed (from outside the AIM realm). One of the ones I like the best is called Walk Forward Testing (search for it in Google for more details). In fact I think it is so valuable that I'm planning on adding it to AI 2.0 as a free update later in the year (it'll be combined with a genetic algorithm to automatically perform optimizations on the various AIM variables -- such as BR, SR, minimum percent per trade, etc.)
At that point, at least AIMers will have a globally accepted testing method in which to compare results.
On another note, the testing I've done over the past year has taken a substantial amount of time (even with the aid of my trusty computer). Late last year/Earlier this year (when I was working on AI 2.0), I implemented a couple of improvements that make testing easier. AI 2.0 includes a multi-analysis function that allows you to enter symbols in a text file and then analyze all of these stocks with one mouse click. (What I do is go to a Web page that lists all of the symbols I'm interested in, e.g. the Dow 30 stocks, and copy them to a text file. Then I use this text file as the input into the multi-analysis function).
Since I've started using this, I've been able to significantly cut down on the time it takes to test. When AI 2.0 is released you can try this function for 30 days free of charge (AI 2.0 alpha testers can try this function now).
Some of the results I've seen (compared to B&H) have really been surprising (especially with some stocks that I had thought were good AIM candidates).
I'm sure that as more people start using this tool in AI 2.0 we'll see many more surprising results.
I'll post more on the subject of testing in the future. If others have testing methods they follow I'd be glad to hear about them.
I'm also curious to know how many people actually backtest the stock(s) they're AIMing. Any comments?
News 
Market Data 
Discover 
