June 1, 2012
After dumping the Ichimoku script and starting to transform a Bollinger Band trading system found on the old Wealth-Lab 4 website, it was obvious that this new system had more potential. At least, it could keep part of its identity.
It turned out that this “new” 2008 system was a variation of a 2002 Bollinger Band system designed by Mark Brown, who participated in one of the forums I visited on LinkedIn. At times, the world may be very small.
Having easily exceeded the Ichimoku Kinko script using the 2008 Bollinger Band system, I wanted to push it a little further. I added new features, one of which, with its collaborative procedures, resulted in buying on the way up and on the way down depending, up to a limit, on the state of the price movement. The outcome was improved performance. It about doubled the preceding performance levels.
All the modifications were applied to AAPL, which served as the testing and debugging price series for the program. Then, I ran once this altered program against the 10 stocks that had already been tested under the last improved Ichimoku Kinko script and under the previous Bollinger Band version that I was modifying.
If the modifications to the previous version were sufficiently general in nature, then most of the 10 stocks in that list would see improved performance levels. And since all 10 stocks covered the same trading interval, then only the trading strategy would explain the differences in behavior and return. Therefore, the improved trading strategy would truly show its advantages over previous versions.
So here are the results of that test, starting with AAPL, which served as the debugging tool:
APPL:
(click to enlarge)
If AAPL was representative of other stocks, in the sense that the trading procedures implemented in this version of the script were of a general nature, then they should apply also to the next 10 stocks as a preliminary test. So, here are those 10 stocks with their respective metrics:
.
(click to enlarge)
.
(click to enlarge)
.
(click to enlarge)
.
(click to enlarge)
.
(click to enlarge)
.
(click to enlarge)
.
(click to enlarge)
.
(click to enlarge)
.
(click to enlarge)
(click to enlarge)
The performance level has increased in all 10 stocks. It should be considered an impossibility as, in the investment industry, there is no such thing as a general-purpose trading strategy. But this is not a general purpose strategy, it is only a set of procedures aimed at reinforcing desirable trading behaviors. If it wants to buy on the way up, it does; if it wants to buy on the way down, it chooses its time and tries to make a small footprint.
The above 11 stocks over the 1,500 trading days (6 years) generated some 10,206 trades with an 85% hit rate. Some 1,489 trades generated losses with an average loss of $1,051. To counterbalance the losses, the average win was $4,286 on 8,717 trades. It resulted in an 81% CAGR over the testing period. Now that is some kind of alpha.
As much as I like this Bollinger Band trading system with its modified behavioral environment, I will let it go. To me, it still generates too many stop losses. I will keep some of the trading procedures as some are very interesting and migrate them to something else that might better suit my personality, which is more of the chicken type.
Applying this improved Bollinger Band system to the rest of the 43 stocks in the SD1 dataset also showed that the improved trading procedures increased overall performance. But still, I want something more.
Created... June 1, 2012, © Guy R. Fleury. All rights reserved