Random stocks experiment
Posted by Danny on August 26, 2013
It doesn’t matter all that much which stocks you buy or when you buy them, what matters most is when you sell them. In a nutshell that’s the conclusion of a stock experiment we have been doing over the last few years. More on that later on in this post, first we will take our weekly look at the stock market.
Markets dipped last week, with many stock indices falling to their lowest levels in more than a month.
We are now entering a new lunar green period, which generally favors rising prices but doesn’t guarantee them.
If the early August high was the peak for a while, then we will now probably get a rather weak green period.
So, we watch carefully how much strength is left in the market. The coming weeks are going to tell us.
Let’s have a look at the S&P 500 chart (click for larger image):
The early August high clearly failed to reach the upper edge of the long term trend channel. That can be seen as sign of weakness and makes it more likely that the bottom of the trend channel will get tested in the current downturn. That would imply a further drop to 1625 or below.
Meanwhile my Earl2 indicator is still dropping rather fast and showing no signs of bottoming out yet. This suggests there is probably more downside action to come. But the shorter term Earl is very oversold and has turned up. The situation is very similar to last June, when we then got a second drop a few weeks later, which painted a bullish divergence in the Earl indicator and marked the low of that correction.
So, even though we are starting a lunar green period, I prefer not to buy at this point because the risk/reward is just not good enough.
I expect better buying opportunities once the Earl2 indicator is bottoming out.
Today I want to wrap up on a little experiment I started in June 2010.
The idea was to equal or better the performance of the S&P 500 index with a portfolio of randomly chosen stocks. The portfolios would use 10 random stocks, and once a month, at the end of lunar green period, the stock that had the worst performance over the preceding month would get sold, putting the portfolio with 10% cash (invested in SHY). Then at the end of lunar red period, the cash would be used to buy a new random stock and bring the portfolio back to fully invested.
This is the basic idea of culling the herd once in while by killing the weakest animal.
These portfolios started doing quite well from the get go, see: Random stock portfolios.
And over the last couple of years they have kept doing quite well. Not beating the S&P 500 by a big margin, but also bear in mind that the portfolios were in 10% cash half of the time, so the result was achieved with lower risk.
Here is the equity curve for the main portfolio, which only used random stocks from the S&P 500 itself:
The portfolio gained 54.2% over the test period, beating the S&P 500 by 0.3%, but with lower risk. You can also see that the equity curve (blue line) has been less volatile than the S&P 500 itself. The second portfolio beat the S&P 500 by 0.7%, so a very similar result.
Perhaps it doesn’t matter all that much which stocks you buy, as long as you don’t hold on to the losers and let the profits run. This is of course achieved by automatically selling off the weakest stock every month. The experiment confirms the importance of cutting losses short and not holding on to bad positions forever. With randomly picked stocks this is easier psychologically, because there is no stock picker’s ego involved. It is fully expected that half of the random picks will not be great stocks, so one can let go of them very easily.
Both portfolios ended with a few stocks that were up almost 100%. Trading on “intuition” one might have taken profits with 20 or 30% gains already, but by only selling the worst stock every month these stocks were kept for much bigger profits. That makes a lot of difference for the longer term performance.
I also noticed that it were often the stocks that rarely make the news that performed very well. These are perhaps stocks that many of us would never buy because nobody talks about them. So, this is another advantage of buying random stocks, you ending up buying some little known gems and they make up for any losses from the bad picks.
Things I would do different for better performance:
Sometimes the randomly picked stock was really bad, e.g. a nearly bankrupt company, and predictably got thrown out after one or two months for a solid loss. I think results could be improved by using a good stock screener to throw out at least the worst of stocks, then making a random pick from the stocks that meet certain minimum criteria.
It may also make sense to use 15 or 20 stocks and remove the two or three worst stocks every month. I noticed that sometimes several stocks started underperforming at the same time, but then it took several months for the portfolio to get rid of them. So, how quickly the bad stocks get flushed out can make a difference, and there must be some optimum.
I am going to stop maintaining these portfolios, as the experiment has served its purpose. If you want to review the experiment, you can find it here: