Way back in 1999 and 2000, just as the .COM bubble was bursting, I started using a mechanical algorithm (RSIBD – Relative Strength IBM) to select stocks for me. I purchased 5 positions, adjusted weekly, and re-balanced anytime one position got about 25% out of line with the others.
Fortuitously, I made some large gains and stopped before the bottom fell out. They key thing I learned is that during a clear Bull trend, momentum strategies can build up some excellent gains – however, they tricky part is making sure you are in a trend that is working for you. Also, you may noticed that the most powerful decision I made, in terms of profit, was when to close my positions. If I had continued using the same strategy, I would have trickled away much of my gains.
A few days ago, I read a mechanical investing article that talked about using stops generated from “average true range” to identify exit points. It occurred to me that this method (if validated) could be used as an indicator in addition to its utility for exits:
- What if all 5 of my positions get “stopped out” within a very short period? Could that indicate a general market downturn?
- What if 80% of my positions do not get “stopped out” for extended periods? Could that indicate a solid positive trend?
- What if I have a lot of turnover or get whipsawed? Market transition, perhaps? Pullback, perhaps?
So, I backtested this method of defining “stops” and tested it starting in January with the SOS-D screen. I only made purchases on Mondays, so if I got stopped out, I stayed cash on that position until the following Monday. Frankly, it was sort of choppy and stopped me out a lot, forcing me to decide if I should buy back into the same position when it came up on the screen again and again. I was making gains though, to the tune of about 20% for the year through May.
Let’s look at a chart of the year for the NASDAQ100 (generated by http://finance.yahoo.com):
What really got interesting though, was when, during the 2nd week of May, I got stopped out of all five positions in a single week. I gave up a little profit, but if I had stayed cash at that point, it would save a huge amount of the market losses through the end of July. If I examine those stops on the chart above, I get a confirmation in the form of the price breaking below the 50 day moving average (and then the 200 DMA, shortly thereafter). I tweaked the methodology by observation as well, adding in a couple of rules for myself to try maximizing the retention of gains when a single position goes bad. I added these two additional alternative stop conditions:
- Close the position on any cumulative position loss of 5%. This means if I happen to buy into a position at a bad time, I will limit my overall position loss to 5%.
- Close the position on any cumulative pullback of 5% from a High Point. This means if I ride a position up to some gains, I get out within 5% of the High.
Staying in cash until the NASDAQ broke back above the 200 DMA would stop a lot of pain. I can use the same system while in cash to monitor and watch for a period when 80% of the positions begin holding for multiple weeks – that *may* identify a time to re-enter that is sooner than the 2000 DMA crossover point, I haven’t tested that yet.
From late July, several of the SOS-D stocked identified have not stopped out up through this week, meaning that the positions would have racked up large gains over the past 3 months. Re-balancing to account for different rates of gain would also help preserve gains if individual positions go bad, though few appear to during this period.
As an additional methology to add to my existing trading techniques, this looks good. I will be opening 3 positions today to put it to the test in real life. I am a little bit nervous taht we may be reaching the end of an uptrend, but at the same time, we’ve broken through the April highs, so there could conceivably be a continued Bull market for a while. And, if I am near the end, at least my new “stop” technique should put me on the side with minimal losses if the trend breaks down.