July 26, 2011

Over last weekend, I ran my modified version of the Momentum Trader (version 2, model 0.7 Level 2). My objective was to try to push performance higher. For a solution, I modified the trend definition, increased the trade unit and jumped to level 2. The performance metrics can be found HERE.

The original version of this script can produce nice results on certain stocks but on others, it should be considered rather dangerous as it has no stop loss. This script has a positive, although awkward, trend definition which can be summarized as: you have 3 up days in a row, then the price is definitely going up. Designing for the long term, it is not sufficient for a trend definition but then again with a few modifications you can coax the script to believe in its new self.

When I run a test of this sort, I usually test the script over one stock. A matter of testing if there are any bugs in the script and to see if it does what I intend it to do. I have to rely on my mental picture of what the script will do over a group of stocks since presently I can only see graphically the last 11 months of the 5.83 years of price data (1,500 bars). I find this acceptable as it forces me to view the whole picture without necessarily viewing all the details. The process, I think, has the advantage of reducing curve fitting and the over-optimization factor usually inherent in our testing procedures.

My version of the Momentum Trader trades a lot, over its 5.83 years lifespan: some 205,000 trades. The need for trade automation becomes self-evident at this level and under these conditions. But then again, why worry, this is what our trading software is designed to do.

Performance wise, my modified version had staggering results. It traded like crazy and showed a 91% hit rate (representing over 187,000 positions). It is not that it was correct in 91% of the trades but that 91% of the trades showed a profit on the last day of trade which was comprised of realized and unrealized profits due to still opened positions. Just as most of the losing positions were still opened positions since stop losses were thought to have been taken relatively early, like within the first 14 weeks after taking a position.

I liked the performance, I liked the numbers and I liked the general behavior of the portfolio as it evolved in time. It is as if the method was gradually reinvesting available excess equity to feed its accumulative process. But wasn't that my intention from the very beginning!


Created on ... July 26, 2011,     © Guy R. Fleury. All rights reserved.