Corrected from original on 4/3/10
I have constructed my systems to generate high winning ratios when tested against large amounts of data. In so doing, I hope to capture trends as well as over-bought and over-sold situations. Our systems were tested with futures data and profitability was optimized in various ways. In the BLOG, we have applied the systems to both futures and to ETFs. The true test of computerized systems is the success rate generated in real-time.
We started publishing results in the BLOG beginning in January. We are happy with the results that have unfolded thus far. One principle used in designing the programs was to cut losses while letting the winning trades “run.” Our biggest success for 2010 is our open Cotton position ETF that is up over 16%. China is now the world’s largest consumer of Cotton, by the way, and China continues to grow economically. Our biggest closed position win was short the Euro Currency (long the U.S. Dollar), with an 11.1% ETF gain. The long dollar trade was not intuitive, at least to me. Right now, the world markets are trying to determine which currencies are the strongest against the backdrop of deficit spending in most of the developed countries. Our biggest loss was 6% on an early cotton trade, and -3.7% on a short Gold trade, which I didn’t like because of gold’s place as a possible “alternative world currency” in these troubled economic times.
Based upon the assumptions below, the return generated by all of the ETF signals in the BLOG was 3.3% for the period January 7, 2010 through March 7, 2010. (Bloomberg reported tonight that the average hedge fund earned .5% in the month of February). We’ll continue to track the return of the BLOG systems as the signals unfold. The open positions could still go up more, or decline in value as the month of March unfolds. We hope that this rate of return will continue at the current pace for the rest of the year, but make no predictions. However, if we annualize the year-to-date return using simple linear extrapolation, it becomes approximately 20%. I would consider this highly unlikely, since real world results almost never follow a straight line. But in any event, we’re off to a good start for the year. We intend to do our next update after the conclusion of the first quarter. Thanks to the three readers that inspired me to design a way of measuring the real-time success of the computer programs.
Assumptions underlying the results above:
1) The returns do not take into account commissions and other trading costs. My bank (which owns a broker) gives me 30 commission-free trades per month. You should take into account your own percentage commission and other trading costs in determining your own returns.
2) The returns do not take into account “slippage,” that is, the phenomenon that the system generated stop may not be realized in reality. The experienced stop may be better or worse than the one generated by the system.
3) On any given day, we do not know how many positions we will have. The number of active positions varies from day to day. Historically, for 2010, it has ranged between one and six, inclusive. For purposes of our calculations, we assumed 5 positions of equal amounts of money. So, for example, if you had $10,000, you would break it into five equal chunks of $2,000. When a signal comes, we assume that only $2,000 is devoted to it, then $2,000 in the next one, etc. So, for example, as much as we wish we would like to have put every dollar in cotton, we assume that only 20% was put into cotton, etc.
4) We assumed no compounding of earnings. In other words, if the portfolio rises in value to $12,000 from $10,000, we still devote only $2,000 in five chunks each time a signal is generated.
5) The yield on un-invested assets is zero. What we mean here is the return on assets not in a system. This yield is pretty realistic for brokerage houses these days. If anything, it is on the conservative side, that is, the returns we state below would be slightly higher if the assets not being used to trade were invested in something with a positive yield.
6) There are a few signals that exited overnight before the opening of the New York markets the next day. I have excluded those signals from the ETF calculation. However, there were a few that exited during the first day of New York trading, and I included those in the results. My computer systems revise their stops continuously, and so there were a few signals that could not have been followed exactly by the readers of the BLOG. Someday, I may offer the facility to send emails out when the systems generate intra-day signals.
7) The results assume that every signal was followed. You should do your own thinking and decide what risk profile you would like to employ. It is very possible that you might have fundamental issues with some of the signals. You may simply be bullish on stocks. Or you might want to be long gold no matter what. Some may wish to get out of a long position when a sell short signal is generated, but not go short. There are all sorts of possibilities for each person. As I said before, the results are illustrative only. You should determine what assumptions are appropriate for you based upon how you use the data presented in the BLOG.
I hope that this line of reasoning is not too complicated. However, portfolio results are only as reliable as the methodology used to bring them about.
Wednesday intra-day update: Canadian ETF stopped out at 97.34; Short Gold stop DGZ 19.51
If you wish to be notified of new posts, let me know at bassanalytics@live.com. I will send you an email every time there is a new post.
Any views expressed herein are provided for information purposes only and should not be construed in any way as an offer, an endorsement, or inducement to invest.
The quotes and symbols used in the BLOG are believed to be reliable, but no guarantees are made with regard to the accuracy.
No comments:
Post a Comment