June 5, 2013

As a follow-up to my previous article on the long-term simulated IBM performance results, I've opted to provide more details on the origin and make-up of the underlying trading strategy. That article concluded with the results of a prior simulation done almost 2 years ago, in July 2011, using almost the same trading script.

IBM was selected in the previous article for the reason that, in 1988, it could have been a reasonable investment choice and could serve as a representative of the group of stocks under simulation.

I stated in a LinkedIn forum on automated trading strategies that:

"Looking at the July 2011 test results, one would expect to have the other members of that portfolio see their respective performance level increase over the 25 years test and therefore show that the trading script would have performed remarkably over the almost 2 years of unseen future data, as well as, the 17 years of data, prior to the original test interval. This will save me the time to have to do such a test again."

After having stated this, I had to run the ADD3 trading script over some of the stocks (other than IBM) that were there some 20-25 years ago. It would immediately show if the script was a one-trick pony or not.

Of the few stocks that had a 20 to 25-year history, all outperformed as expected. In a way, it's not helping my case. It again falls into the incredible realm: the too-good-to-be-true kind of thing. Yet, these tests are just simulations of a specific trading script over a series of stocks. The ADD3 scripts were put aside mainly because they had no "finesse". It was like having used a chainsaw to structure the script. It also lacked some of what would become my preferred trading procedures and safeguards.

Nonetheless, here are some of the stocks in my tests that were there some 20 or more years ago: 

ADM

ADM

(click to enlarge) 

AMZN

AMZN

(click to enlarge) 

AAPL

AAPL

(click to enlarge)

CAT

CAT

(click to enlarge)

CERN

CERN

(click to enlarge)

CSX

CSX

(click to enlarge)

DECK

DECK

(click to enlarge)

IDCC

IDCC

(click to enlarge)

PCLN

PCLN

(click to enlarge)

PNRA

PNRA

(click to enlarge)

The ADD3 Script

The ADD3 script tries to accumulate shares over time and will trade over the accumulative process. It will feed back the proceeds of trading to acquire more shares over the trading interval. The concept is simple. Its implementation is something else.

It is always delicate to show simulation results like these. One of my major concerns is the credibility factor. However, I do think that anyone could have programmed a script like this, and I hope that maybe, soon, someone will. And it will be his/her unique solution to the problem.

When you look at the metrics of these performance summary reports, the first thing that stands out is that each individual stock had its own signature, no two reports are the same; as should have been expected. And yes, the biggest drawdown was most often the one to come. This is easily understandable, as the portfolio grows with its growing inventory, so is the dollar amount of any dip or retracement.

This all restates what has been said before. A trading strategy, even roughly designed as in this case, can be effective over long trading intervals. There are trading strategies that do not fall down when applied to unseen data at both ends of the trading interval. And maybe, most importantly, trading over a stock accumulation process does have merit after all.

The ADD3 program was designed almost blind, in the sense that only part of the data was visible, some 220 trading days out of 1,500 (about 11 months out of 6 years). When viewed from a long-term perspective of 25 years, as tested above, this would have represented 220 days of seen data over 6,250 trading days. There were over 450 trading days of unseen future data, which showed a sufficiently long time interval where the program did not fall apart; while there were some 4,250 trading days of data prior to the initial testing period itself, another place where the program should have broken down but did not.

The point I am making is that if the trading methodology is sound over even a small window of viewable data and the trading principles applied are designed to stand the test of time, then the trading strategy can be extended forward or backward without breaking down.

For me, what I saw in these few long-term tests was that, indeed, the CAGR tapers off with time as it should. This was expected, as expressed in my first papers. My explanations for this can be found here

There was some inefficiency in the ADD3 script as it tends to also accumulate cash, which reduces market exposure, which in turn gradually reduces long-term performance. This is why I went on to design other better and "improved" trading scripts.


 Created... June 12, 2013,    © Guy R. Fleury. All rights reserved.